AutoUni – Schriftenreihe

Chris David Gabrisch

Eco-Effectiveness
of Modular Products
and Fleets within
the Automotive
Industry
AutoUni – Schriftenreihe
Volume 164

Reihe herausgegeben von

Volkswagen Aktiengesellschaft, Volkswagen Group Academy, Volkswagen
Aktiengesellschaft, Wolfsburg, Germany
Chris David Gabrisch

Eco-Effectiveness
of Modular Products
and Fleets
within the Automotive
Industry
Chris David Gabrisch
Salzgitter, Germany

Admitted dissertation of the Technical University Carolo-Wilhelmina zu Braunschweig,

2022.

The results, opinions and conclusions of the AutoUni publication series published
doctoral theses are only those of the doctoral candidate.

ISSN 1867-3635 ISSN 2512-1154 (electronic)

AutoUni – Schriftenreihe
ISBN 978-3-658-40593-9 ISBN 978-3-658-40594-6 (eBook)
https://doi.org/10.1007/978-3-658-40594-6

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer
Fachmedien Wiesbaden GmbH, part of Springer Nature 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher,
whether the whole or part of the material is concerned, specifically the rights of translation, reprint-
ing, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other
physical way, and transmission or information storage and retrieval, electronic adaptation, computer
software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with
regard to jurisdictional claims in published maps and institutional affiliations.

This Springer Vieweg imprint is published by the registered company Springer Fachmedien
Wiesbaden GmbH, part of Springer Nature.
The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany
Disclaimer

Ergebnisse, Meinungen und Schlüsse dieser Dissertation/Veröffentlichung sind

nicht notwendigerweise die der Volkswagen Aktiengesellschaft.
The results, opinions and conclusions expressed in this thesis are not
necessarily those of Volkswagen Aktiengesellschaft.

v
Abstract

The automotive industry is facing the challenge of reducing its environmen-

tal impact to comply with stricter fleet emission regulations across all relevant
markets. Thus, new technologies, e.g. lightweight materials, or an electrified pow-
ertrain need to be integrated into the new products to reach the reduced emission
limits for the products. Many innovations that reduce emissions during the use
stage of a vehicle often come with higher burdens during the production stage or
at the end of life stage. However, both these stages are currently not regulated.
The current emission targets are set for fleet averages in the use stage only. Thus,
the OEMs need to find the ideal product configuration based on new technolo-
gies and design the ideal fleet composition to reach this target with the least cost
effort. Still, an OEM’s contribution to the targets of the Paris Agreement must
consider the entire life cycle of a vehicle, surpassing the targets of the current
legislations which focusses on the use stage only.
This work presents a concept that identifies the ideal configuration of a mod-
ular product system like a vehicle to meet a limited environmental impact at the
lowest life cycle costs along the entire life cycle. This optimization is based on the
ideal combination of modular product components which are selected by an algo-
rithm based on graph theory. A modified shortest path algorithm is used to define
the flow through a network of available components with individual weights for
costs and emission on every arc. Expanded to a minimum-cost flow problem,
multiple agents, e.g. representing a fleet of vehicles, are sent through the network
with the target to identify the ideal fleet configuration. Herein lies the challenge
of handling limited availabilities of certain options (e.g. scarce green electric-
ity for the use stage of electric vehicles). The presented methodology includes a
mechanism that assigns such limited options to those individual vehicles within
a fleet that profit the most from that specific option.

vii
viii Abstract

The application to a case study based on a Volkswagen Golf VII model and an
exemplary fleet of 100,000 vehicles of this type of vehicle shows the advantages
of the developed concept. Four different emission limitation scenarios, ranging
from 35t CO2 eq. to 21t CO2 eq. over the entire life cycle, have been assessed
for both the single vehicle and the vehicle fleet. The results show that for fleet
targets within the defined assumptions and scenarios, a fleet-based optimization
for 100,000 vehicles is more efficient compared to a single vehicle optimization
which is then rolled out to 100,000 vehicles. For relatively simple emission stan-
dards, the fleet is ideally based on combustion engines with optimized driving
resistances, e.g. by using lightweight materials. For stricter standards, electrifi-
cation is indispensable. Given the defined assumptions and scenarios, a tipping
point is reached at an emission limitation of 28t CO2 eq. per vehicles life time.
Starting from this emission limit, the algorithm stops using gasoline powered
vehicles and starts to incorporate electric vehicles into the fleet composition.

Keywords Life cycle assessment · Life cycle costing · Graph theory · Eco-
effectiveness · Fleet emission reduction
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Present Situation and Problem Statement . . . . . . . . . . . . . . . . . . . . 1
1.2 Research Objective and Structure of Thesis . . . . . . . . . . . . . . . . . . 5
2 Theoretical Background and Technical Overview . . . . . . . . . . . . . . . . 9
2.1 Modularity of Product Systems and Product System Fleets . . . . . 9
2.1.1 Modular Product System Architectures . . . . . . . . . . . . . . . . 9
2.1.2 Multiple Product Systems and Population Fleets . . . . . . . . 18
2.2 Life Cycle Thinking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.1 Environmental Life Cycle Assessment . . . . . . . . . . . . . . . . 22
2.2.2 Life Cycle Costing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Sustainable Development and Corresponding Legislation . . . . . . . 33
2.3.1 Environmental Impacts and Impact Assessment . . . . . . . . 33
2.3.2 Sustainability and Sustainable Development . . . . . . . . . . . 36
2.3.3 General Environmental Legislation . . . . . . . . . . . . . . . . . . . 41
2.3.4 Environmental Regulations for
the Automotive Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Life Cycle Engineering and Mathematical Optimization . . . . . . . 51
2.4.1 Life Cycle Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.4.2 Optimization Approaches for Decision Support . . . . . . . . 57
2.5 Conclusions Regarding the Theoretical Background . . . . . . . . . . . 60
3 State of Research and Identification of the Research Gap . . . . . . . . . 63
3.1 Criteria and Requirements for Optimizing
the Eco-effectiveness of Product Systems . . . . . . . . . . . . . . . . . . . . 63
3.1.1 Criteria to Handle Product System Modularity . . . . . . . . . 64
3.1.2 Criteria to Handle Use Case Specific Requirements . . . . . 65

ix
x Contents

3.1.3 Criteria for Optimization Approach . . . . . . . . . . . . . . . . . . . 66

3.1.4 Additional Requirements to Obtain Useful Results . . . . . . 67
3.2 Current Approaches and State of Research . . . . . . . . . . . . . . . . . . . 68
3.2.1 Product Modularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.2 Modular Life Cycle Assessments . . . . . . . . . . . . . . . . . . . . . 72
3.2.3 Optimization of LCA and LCC . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Identification of the Research Gap . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4 Concept for the Optimization of Eco-effectiveness of Product
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.1 Concept Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.2 Framework for the Optimization Concept . . . . . . . . . . . . . . . . . . . . 93
4.2.1 General Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2.2 Selection of Graph Theory as Optimization Approach . . . 96
4.3 Modelling the Product System’s Structure in a Network . . . . . . . 103
4.3.1 Transformation of Modular Product Systems
into Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.3.2 Product System Networks Including
Interdependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.3.3 Network Reduction Strategies for Interdependency
Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.3.4 Strategy Adaption to Reduce the Data Demand
of LCA Values and LCC Values . . . . . . . . . . . . . . . . . . . . . . 115
4.4 Data Management of the Input Data . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.5 Adaption of Shortest Path Algorithms to the Problem
Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.6 Visualization and Interpretation of Results . . . . . . . . . . . . . . . . . . . 132
5 Prototypical Implementation and Application of the
Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.1 Prototypical Implementation of the Optimization Approach . . . . 137
5.2 Exemplary Application Cycle of the Optimization Approach . . . 141
6 Application of the Optimization Approach to a Case Study
of the Automotive Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.1 Life Cycle Perspectives of a Vehicle . . . . . . . . . . . . . . . . . . . . . . . . 146
6.1.1 Product Life Cycle of a Vehicle . . . . . . . . . . . . . . . . . . . . . . 146
6.1.2 Environmental Assessment of a Vehicle’s Life Cycle . . . . 148
6.1.3 Total Cost of Ownership along a Vehicle’s Life
Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Contents xi

6.2 Selection of Measures for the Reduction of Greenhouse

Gas Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.2.1 Selection of Measures and Module Alternatives . . . . . . . . 156
6.2.2 Measure Analysis Regarding LCA, LCC and
Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.3 Data Input for the Vehicle and Fleet Optimization . . . . . . . . . . . . 168
6.4 Results of Optimization for Different Scenarios . . . . . . . . . . . . . . . 171
6.4.1 Analysis of the Results for a Single Vehicle
Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.4.2 Analysis of Results for Vehicle Fleet Optimization . . . . . 177
6.4.3 Sensitivity Analysis of Results . . . . . . . . . . . . . . . . . . . . . . . 184
6.5 Findings for Further Vehicle Development and Fleet
Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
7 Summary, Critical Appraisal and Outlook . . . . . . . . . . . . . . . . . . . . . . . 193
7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
7.2 Critical Appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.3 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Abbreviations

AP Acidification potential
AbwAG Abwasserabgabengesetz
BEV Battery electric vehicle
BimSchG Bundes-Immissionsschutzgesetz
CAFC Corporate average fuel consumption
CARB California air resources board
CCT Computational complexity theory
ChemG Chemikaliengesetz
CNG Compressed natural gas
CO Carbon monoxide
CO2 Carbon dioxide
CO2 eq. Carbon dioxide equivalents
eGas Regenerative compressed natural gas
EPA Environmental protection agency
EU European Union
EU ETS European Union emissions trading system
FRV Fuel reduction value
FTP-75 Federal test procedure 75
GDP Gross domestic product
GHG Greenhouse gases
GWP Global warming potential
HC Hydrocarbons
IEA International energy agency
LCA Life cycle assessment
LCC Life cycle costing
LCE Life cycle engineering

xiii
xiv Abbreviations

LCI Life cycle inventory

LCIA Life cycle impact assessment
LCM Life cycle management
MCFP Minimum-cost flow problem
NEDC New European driving cycle
NMHC Non-methane hydrocarbons
NOX Oxides of nitrogen
NP Nondeterministic polynomial time
ODP Ozone depletion potential
OEM Original equipment manufacturer
PCOF Photochemical ozone formation
PHEV Plug-in hybrid electric vehicle
PM Particular matter
PN Number of particular matter
TCO Total cost of ownership
UN United Nations
UNEP United Nations environment programme
USA United States of America
VDI Verein Deutscher Ingenieure
WBCSD World business council for sustainable development
WCED World commission on environment and development
WLTP World harmonized light duty test procedure
ZLEV Zero and low emission vehicles
List of Figures

Figure 1.1 Perspectives on regulation of product fleets . . . . . . . . . . . . . 4

Figure 1.2 General structure of the dissertation . . . . . . . . . . . . . . . . . . . 7
Figure 2.1 Visualization of product architecture . . . . . . . . . . . . . . . . . . . 11
Figure 2.2 Definition of a modular product system structure . . . . . . . . 13
Figure 2.3 Modularity of product units . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Figure 2.4 Hierarchical decomposition of a product . . . . . . . . . . . . . . . 16
Figure 2.5 Network of product
components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Figure 2.6 Fleet of multiple product systems . . . . . . . . . . . . . . . . . . . . . 21
Figure 2.7 Flow-oriented life cycle model . . . . . . . . . . . . . . . . . . . . . . . 23
Figure 2.8 Life cycle assessment
framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Figure 2.9 Processes with input and
output flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Figure 2.10 Life cycle impact assessment . . . . . . . . . . . . . . . . . . . . . . . . . 26
Figure 2.11 Classification and characterization of the
impact assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Figure 2.12 Life cycle costing by the VDI . . . . . . . . . . . . . . . . . . . . . . . . 30
Figure 2.13 Fixed and variable
costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Figure 2.14 Life cycle costing—stages and costs . . . . . . . . . . . . . . . . . . . 33
Figure 2.15 Chain of effects for environmental impacts . . . . . . . . . . . . . 34
Figure 2.16 Life cycle impact assessment midpoint framework . . . . . . . 35
Figure 2.17 Shares of greenhouse gases and main causes . . . . . . . . . . . . 37
Figure 2.18 Depiction of the triple bottom line . . . . . . . . . . . . . . . . . . . . 38
Figure 2.19 Eco-efficiency vs. eco-effectiveness . . . . . . . . . . . . . . . . . . . . 40

xv
xvi List of Figures

Figure 2.20 Defining an environmental target . . . . . . . . . . . . . . . . . . . . . . 42

Figure 2.21 Transport-related CO2 emissions
in the European Union . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Figure 2.22 Comparison of NEDC and
WLTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Figure 2.23 Comparison of CO2 emission limitations worldwide . . . . . 51
Figure 2.24 The product design
paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Figure 2.25 Life cycle engineering framework . . . . . . . . . . . . . . . . . . . . . 54
Figure 2.26 Eco-effectiveness and life cycle engineering . . . . . . . . . . . . 56
Figure 2.27 Categorization of problem types . . . . . . . . . . . . . . . . . . . . . . 59
Figure 3.1 Focus area for literature review . . . . . . . . . . . . . . . . . . . . . . . 69
Figure 3.2 Liaison graph for module connections . . . . . . . . . . . . . . . . . 71
Figure 3.3 Comparison of conventional LCA and modular LCA . . . . . 73
Figure 3.4 Representation of alternative module
combination options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Figure 3.5 Design alternatives for product design . . . . . . . . . . . . . . . . . 81
Figure 4.1 Framework for the optimization of eco-effectiveness
for modular product systems . . . . . . . . . . . . . . . . . . . . . . . . . 95
Figure 4.2 Functionality of the user interface . . . . . . . . . . . . . . . . . . . . . 97
Figure 4.3 Undirected graph and undirected and induced subgraph . . . 98
Figure 4.4 Weighted and directed graph and the shortest path . . . . . . . 99
Figure 4.5 Network for minimum-cost flow problem . . . . . . . . . . . . . . . 99
Figure 4.6 Integration of additional nodes . . . . . . . . . . . . . . . . . . . . . . . . 101
Figure 4.7 Multiple limitations for different life cycle stages . . . . . . . . 103
Figure 4.8 Modular product unit as a network . . . . . . . . . . . . . . . . . . . . 105
Figure 4.9 Product system structure network of a modular
product system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Figure 4.10 Structure networks with interdependencies . . . . . . . . . . . . . . 109
Figure 4.11 Network B with a separated structure network . . . . . . . . . . 112
Figure 4.12 Network structure with correction factor . . . . . . . . . . . . . . . 114
Figure 4.13 Separation of use stage from product system
network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Figure 4.14 Difference between node demand and data demand . . . . . . 120
Figure 4.15 Applying reduction strategies to a system structure
network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Figure 4.16 Flow chart of the general logic of the optimization
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Figure 4.17 Definition of key variables and arc design . . . . . . . . . . . . . . 126
List of Figures xvii

Figure 4.18 Definition of arc design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Figure 4.19 Definition of optimization objectives . . . . . . . . . . . . . . . . . . . 128
Figure 4.20 Definition of constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Figure 4.21 Exemplary provision of input data (I) . . . . . . . . . . . . . . . . . . 130
Figure 4.22 Exemplary provision of input data (II) . . . . . . . . . . . . . . . . . 131
Figure 4.23 Export of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Figure 4.24 Visualization of results in the network . . . . . . . . . . . . . . . . . 133
Figure 4.25 Scatter plot for visualization of results . . . . . . . . . . . . . . . . . 134
Figure 4.26 Scatter plot visualization of results for single product
optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Figure 4.27 Scatter plot visualization of results for multiple
product optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Figure 5.1 Summary of implemented software tools to cover all
required steps of the approach . . . . . . . . . . . . . . . . . . . . . . . . 138
Figure 5.2 Exemplary user interface for data provision
to the optimization algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 140
Figure 6.1 Life cycle of a vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Figure 6.2 Share of materials within a vehicle . . . . . . . . . . . . . . . . . . . . 147
Figure 6.3 End of life treatment of vehicles . . . . . . . . . . . . . . . . . . . . . . 149
Figure 6.4 Greenhouse gas emissions of different
types of powertrains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Figure 6.5 Distribution of emission sources
along the life cycle stages . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Figure 6.6 Total costs of ownership from
customer’s perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Figure 6.7 Distribution of the TCO along the different life cycle
stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Figure 6.8 Possible emission reduction approaches . . . . . . . . . . . . . . . . 158
Figure 6.9 Module network for case study application . . . . . . . . . . . . . 160
Figure 6.10 Network logic for optimization network . . . . . . . . . . . . . . . . 170
Figure 6.11 First 120 lines of data structure table . . . . . . . . . . . . . . . . . . 172
Figure 6.12 Emission and TCO profile of all vehicle
configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Figure 6.13 Network flow for optimized vehicles . . . . . . . . . . . . . . . . . . 176
Figure 6.14 Scatter plot results for single vehicle optimization . . . . . . . 178
Figure 6.15 Vehicle fleet optimization for scenario one . . . . . . . . . . . . . . 179
Figure 6.16 Vehicle fleet optimization for scenario two . . . . . . . . . . . . . 179
Figure 6.17 Network flow for fleet optimization scenario two . . . . . . . . 180
Figure 6.18 Vehicle fleet optimization for scenario three . . . . . . . . . . . . 181
xviii List of Figures

Figure 6.19 Network flow for fleet optimization scenario three . . . . . . . 181
Figure 6.20 Vehicle fleet optimization for scenario three with use
stage limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Figure 6.21 Vehicle fleet optimization for scenario four . . . . . . . . . . . . . 183
Figure 6.22 Network flow for fleet optimization scenario four . . . . . . . . 184
Figure 6.23 Results of sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . 186
Figure 6.24 Fleet composition for sensitivity analysis scenarios . . . . . . 187
Figure 6.25 Comparison of vehicle and fleet results . . . . . . . . . . . . . . . . 188
Figure 6.26 Result overview for all fleet scenarios . . . . . . . . . . . . . . . . . 190
Figure 6.27 Share of powertrains for all emission limit scenarios . . . . . 191
List of Tables

Table 2.1 Selected environmental impact categories, indicators,

and reference units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Table 2.2 Limitation values for air pollutants
set by Euro Standard 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Table 2.3 Complexity classes and big O notation . . . . . . . . . . . . . . . . . . 60
Table 3.1 Criteria for optimization of eco-effectiveness
for product systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Table 3.2 Evaluation of published approaches for LCA and LCC
optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Table 4.1 Data structure table for an exemplary network structure . . . . 123
Table 6.1 Purchasing costs of different powertrains . . . . . . . . . . . . . . . . 153
Table 6.2 Assumed costs of use stage for different powertrains . . . . . . 154
Table 6.3 Summary of input data for basic vehicles and energy
carriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Table 6.4 Input data for production-focused alternatives . . . . . . . . . . . . 164
Table 6.5 Input data for use stage-focused alternatives . . . . . . . . . . . . . . 165
Table 6.6 Single vehicle optimization results . . . . . . . . . . . . . . . . . . . . . . 175
Table A.1 Data structure table with algorithm input data . . . . . . . . . . . . 200

xix
Introduction
1

1.1 Present Situation and Problem Statement

Driven by an increasing global population, economic globalization, and higher

standards of living in developing countries, the global primary energy demand has
rapidly grown in the past (Herrmann 2010, p. 35). For the future, the International
Energy Agency (IEA) predicts a continuation of this trend. The primary energy
demand is supposed to increase by more than 30% towards 207.78 billion MWh
until 2040 compared to the level of 2014 (IEA 2016, p. 61). Along with the
rising demand for energy, the global exploitation of resources and raw materials
such as biomass, metals and nonmetallic minerals is supposed to rise from 88.6
billion tons in 2017 to over 180 billion tons in 2050, following the current trend
(UNEP IRP 2017). These human-made interactions with the earth’s environment
are accompanied by numerous direct and indirect consequences. The extraction
of resources and the utilization of various energy forms impact the air, soil, and
water of the earth’s ecosystem through exhaust gases and wastes (Herrmann 2010,
p. 13). The influence of such an extensive use of energy and resources on the
ecosystem further leads to negative environmental effects like human or eco-
toxicity, ozone depletion, climate change or land use impacts (Jolliet et al. 2004,
p. 395).
Different political organizations have ratified binding agreements to strengthen
environmental protection or to limit environmental pollution, aiming to mitigate
negative environmental effects. The Paris Agreement of 2015 is one example.
195 signatories have agreed to “holding the increase in the global average tem-
perature to well below 2 °C above pre-industrial levels and pursuing efforts to limit
the temperature increase to 1.5 °C above pre-industrial levels” (United Nations

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 1

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_1
2 1 Introduction

2015a, p. 3). In the seventh environment action program of the European Union
(EU), the participating countries agreed to reduce the greenhouse gas emissions
by 20% and to increase the share of renewable energies as well as the energy
efficiency by 20% by 2020 (European Parliament 2013, p. 2). Furthermore, also
customers start to demand products with less negative environmental impacts and
a more ecologically friendly image (Cerri, Taisch & Terzi 2014, p. 1). Hence,
reducing the environmental footprint of a product is also a competitive advantage
for companies.
Companies need to apply measures for the impact reduction already within the
early product development phase to meet such environmental targets (Herrmann
2010). It is also important to evaluate the environmental performance of a product
along the entire life cycle to reduce the environmental impacts (e.g. greenhouse
gas emissions) in a holistic manner and to avoid the shifting of environmental
burdens from one life cycle stage to another. Consequently, the environmental
performance of a product needs to be measured in line with the principles of
Life Cycle Assessments (LCA) (DIN ISO 14040).
Considering the entire life cycle of a complex product, there are different mea-
sures or options that change the product system and help to reduce environmental
burdens like the emissions of greenhouse gases (GHG) of the product. Measures
that influence and improve the environmental performance of a product along its
entire life cycle are e.g. utilizing recycled materials, reducing the energy con-
sumption of the product, or increasing the energy efficiency as well as utilizing
renewable energy sources (IPCC 2014). For many product types, such measures
are implemented into a product’s life cycle in many different forms and during
different stages along the life cycle. All measures with the purpose of reducing
e.g. the GHG emissions of a product can be characterized and distinguished by
differences regarding their impact power, time horizons, scalability, availability,
interdependencies with other measures as well as cost structures.
Companies that produce goods for customers in an open market operate within
a field of economic competition. Their decisions regarding product changes there-
fore need to consider financial consequences (Paul 2015). New environmental
targets concerning a specific product require changes to the product system which
should therefore ideally be realized with least financial extra costs. Thus, it is
possible for trade-offs between financial and environmental targets to occur that
require an eco-efficient solution (Herrmann 2010, p. 47). Following this, an envi-
ronmental target of a product should be achieved by choosing those measures
that yield the most economically efficient realization.
For products with a complex life cycle and a modular product architecture,
many measures exist to alter the environmental performance. A modular product
1.1 Present Situation and Problem Statement 3

architecture is defined by individual and independent modules that are physi-

cally detached from or can be combined with each other via defined interfaces
(Gu & Sosale 1999, p. 387 ff.). Targets for modular products can be realized
by e.g. replacing one module with an alternative that provides a better environ-
mental performance. Exploring alternative product module options consequently
helps to optimize entire product systems. However, the number of theoretically
possible configurations for a modular product quickly reaches an extremely high
number. Each of the individual configurations provides individual financial and
environmental performances over the product’s lifetime, adding more complexity.
The complexity is even further increased if not only a single product unit is the
target of regulation but an entire population, also called a fleet, of products. In the
automotive context, such a fleet can be the yearly production of an original equip-
ment manufacturer (OEM), but can also refer to all vehicles within a city or the
fleet of a company. As every single product unit within such a fleet can be con-
figured individually, many different fleet configurations are possible. Figure 1.1
shows multiple product units which together form a fleet of individual products.
Such a product fleet is regulated from different perspectives following different
environmental targets that need to be fulfilled. If the regulation (or an OEM with
self-set goals) defines targets from a bottom-up perspective, every product unit
within the fleet needs to meet these targets individually. In a top-down perspec-
tive, targets are set for the entire fleet in general. Here, the target must be fulfilled
by all product units together, e.g. as an average result. This means that individual
products can exceed the limit if other products remain below the limit. While the
automotive industry is currently regulated regarding the use stage emissions only,
the legislation can be expanded to cover the entire life cycle in future. Hence, the
product development strongly depends on the given perspectives and targets that
the product units and fleet must meet.
Modular products in fleets quickly scale up to enormous combinatorial pos-
sibilities due to their nature of exponential growth. As equation 1.1 shows, the
number of possible combinations (c) depends on the number of alternatives per
module (a), the number of modules within a product unit (m) and the number of
product units within a fleet (n).

c = a mn (1.1)

Following equation 1.1, an exemplary product that consists of only three modules
with each module offering three alternatives and a product fleet of five product
units, a total of over 14 million possible fleet configurations are possible. Con-
sequently, the identification of the ideal set of measures for each product unit
4 1 Introduction

regulaon of product fleets

perspecve
“top down”
fleet

overall targets for fleets

(e.g. CO2 compliant fleet)
target must be fulfilled as
average over all product units
perspecve
“boom up”

specific targets per product unit

unit

(e.g. best in class)

target must be fulfilled by
every product unit of the fleet

life cycle regulaon

perspecve

targets defined for the enre life cycle

life cycle

material extracon producon use phase end of life

use phase regulaon

targets for a single stage

Figure 1.1 Perspectives on regulation of product fleets

of a product given specific targets is a complex task that requires a systematic

approach. Within this work, a method for calculating eco-effective pathways in
order to achieve an environmental target for a single product or multiple prod-
uct units by meeting an absolute environmental target with an optimized cost
structure is developed.
One of the biggest sources for GHG emissions is the transport sector and
herein the automotive industry. In Europe, over 13% of the emissions are
caused by road-based passenger traffic (European Parliament 2019). While other
industries were able to reduce their emission level over the past years, the
transportation sector has increased its emissions due to rapidly growing trans-
port demand. The European Union therefore has set strict regulations for the
automotive industry to drastically reduce its climate impact by 2050 (European
Parliament 2019). Since 2020, the limit for CO2 emissions for an OEM’s fleet
is set to 95 grams of CO2 per kilometer on average. This limit will be further
reduced by −5% in 2025 and −37.5% in 2030. (European Parliament 2019) This
demonstrates how urgently the automotive industry needs to identify solutions to
comply with new emission standards in order to avoid potential penalties.
1.2 Research Objective and Structure of Thesis 5

Considering the automotive sector, strict targets for emission reduction

demand extensive changes within the industry, as new solutions require new tech-
nologies that need to be integrated into the fleets in a short time span. Regarding
the high relevance of the automotive industry for climate related issues and the
ongoing technological change process in this industry, the method developed in
this work will be closely designed for an automotive application and validated
with a transfer to an example of a vehicle’s life cycle.

1.2 Research Objective and Structure of Thesis

The target of this work is to enable decision makers to select combinations of

measures or module alternatives for modular product configurations that fulfill
environmental targets with least costs. An analytical evaluation of possible mea-
sure combinations aims to disclose hidden potentials of individual measures on
the environmental footprint of a product along the entire life cycle. Also, mis-
conceptions of individual measures due to their performance within the whole
product system are revealed. Furthermore, this work shall also support invest-
ment decisions for optimizing the environmental performance. Given the case
that a company voluntarily plans to invest into the optimization of their products
environmental footprint, the developed approach provides recommendations on
measures that result in highest environmental benefits for the available budget.
Within this work, a methodology is developed that systematically evaluates all
technically and logically possible configurations of alternative product systems
regarding their financial and environmental life cycle performance. Addition-
ally, interdependencies between combined measures as well as external financial
impacts, like varying material prices, are considered. Based on the data of the
financial and ecological performance, an optimization algorithm is applied that,
for an environmental target or financial budget, identifies and displays the most
effective set of measures or module combinations for the product system or fleet
of multiple product systems.
The presented methodology is applicable to problems considering the config-
uration of just one product system, but can also be expanded to problems where
the given target is set as the average for a group of multiple product systems.
An example for the latter problem type is the limitation of carbon dioxide (CO2 )
emissions for the fleet of car manufacturers. While individual vehicles can be
above the limit, the average fleet emission must meet the restriction. (European
Parliament 2009)
6 1 Introduction

In detail, the objectives of this thesis are:

• to provide decision support for decision makers in the automotive industry

regarding the cost-effective compliance with environmental targets,
• to develop a methodology that identifies the most eco-efficient product con-
figuration while considering interdependencies between measures along the
entire life cycle, and
• to enable the application to not only single but also multiple products of the
same type as e.g. a vehicle fleet.

This work is structured in seven consecutive chapters (see figure 1.2). The second
chapter introduces the reader to the theoretical background of the covered topic.
Preliminary remarks and definitions of modular product systems are presented in
subchapter 2.1. Afterwards, an introduction to the concept of product life cycles
and their evaluation is given. Established procedures to describe the borders and
individual stages of a product’s life cycle are presented, as well as methods to
assess the ecological (life cycle assessment) and financial (life cycle costing)
performance of a product (subchapter 2.2). Then, the theoretical background
of sustainability and the current situation of general and automotive-specific
legislation regarding environmental impacts is explained in subchapter 2.3. In
subchapter 2.4, life cycle engineering and optimization methods are introduced.
The third chapter introduces the challenges and requirements that need to be
fulfilled to identify optimized eco-effective product systems and criteria that are
required to evaluate the suitability of the developed approach (subchapter 3.1).
Based on a literature review, the current approaches within this field of study are
analyzed and an overview of the present state of research for the optimization of
a product’s eco-effectiveness is provided (subchapter 3.2). The state of research is
then evaluated regarding the criteria from subchapter 3.1 and the present research
gap is highlighted (subchapter 3.3).
In the fourth chapter, the requirements derived from the research gap are
discussed (subchapter 4.1), followed by a detailed description of the newly
developed framework and the selected optimization approach (subchapter 4.2).
The modelling of the product systems and the provision and preparation of
the required input data are elaborated in subchapter 4.3. Chapter four contin-
ues with the description of the developed data management (subchapter 4.4) and
the adaption of the selected optimization approach to the problem statement (sub-
chapter 4.5). Finally, chapter four closes with the visualization and interpretation
of the output of the optimization (subchapter 4.5).
1.2 Research Objective and Structure of Thesis 7

1. Introducon
movaon, research queson & objecves

2. Theorecal background
preliminary remarks & technical overview

Modular Life cycle Sustainability & Life cycle

2.1 2.2 2.3 2.4
product systems thinking legislaon engineering

3. State of research
challenges, current approaches and need for acon

Challenges and Current Idenficaon of

3.1 3.2 3.3
criteria approaches research gap

4. Development of methodology
Framework, input-data, opmizaon approach, interpretaon

4.1 / Requirements & 4.3 / Modelling &

4.5
Opmizaon Visualizaon &
4.6
4.2 framework 4.4 input data algorithm interpretaon

5. Prototypical implementaon
Exemplary implementaon and applicaon cycle of methodology

Prototypical Applicaon
5.1 5.2
implementaon cycle

6. Exemplary applicaon of methodology

Applicaon to a problem statement of the automove industry

Life cycle of 6.2 / Measures and 6.4 / Results &

6.1
automobiles 6.3 input data 6.5 interpretaon

7. Summary & outlook

summary, crical acclaim & outlook

Figure 1.2 General structure of the dissertation

Chapter five describes a prototypical implementation of the developed method-

ology using selected tools for the individual steps to present one feasible way for
a technical application of the developed methodology (subchapter 5.1). Also, an
exemplary application cycle is described to show how the tool must be operated
to receive the desired output (subchapter 5.2).
Chapter six discusses the exemplary application of the methodology to a cur-
rent problem within the automotive industry. In subchapter 6.1, the typical life
cycle of a vehicle (based on the concepts of Section 2.2) is described, followed
8 1 Introduction

by a selection and evaluation of applicable measures for improving the environ-

mental performance of a vehicle and the preparation of the specific input data
(subchapter 6.2 and 6.3). The subchapters 6.4 and 6.5 present the results of the
calculations and derive recommendations by interpreting the results.
The seventh chapter finally closes with a summary, a critical appraisal of the
approach and an outlook on further fields of study.
Theoretical Background and Technical
Overview 2

In order to identify the ideal product system configuration for single or multiple
products, several fundamental terms and definitions need to be discussed. This
chapter introduces the fundamentals regarding the structure of modular prod-
ucts, the life cycle perspective, aspects of sustainable development, environmental
legislation as well as life cycle engineering and the principles of mathematical
optimization.

2.1 Modularity of Product Systems and Product System

Fleets

Products and their modular architecture are the foundation of the environmen-
tal optimization along the entire life cycle of product systems. In the following
sections, the definition of products, product systems and a modular product archi-
tecture are discussed. Furthermore, product families and product fleets are also
defined. Lastly, it is shown how the step from a single product unit to a popula-
tion of multiple units changes the complexity regarding available configurations
of product families.

2.1.1 Modular Product System Architectures

A product is defined as the result of processes or activities. Products are distin-

guished between physical goods, software, knowledge, information, and services.

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 9

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_2
10 2 Theoretical Background and Technical Overview

A product can also be a combination of these forms. (Müller 2013, p. 10) The
DIN ISO 14040 for environmental management and life cycle assessment defines
a product as “any good or service” and distinguishes between services, software,
hardware, or processed materials (DIN ISO 14040, p. 8).
The term “production” focuses on the human caused transformation of objects
regarding changes in quantity, quality, spatial or temporal aspects. While temporal
and spatial changes refer to logistics, the changes of quantity and quality strive
for the creation of new goods or services, leading to an added value. This makes
the production a core process of every company. (Dyckhoff & Spengler 2007,
p. 3)
Matiz-Rubio et al. (2020) define a product system as the entire life cycle
from resource provision over the production and use stage to the end of life
of the product, including co-products (Matiz-Rubio, Eltrop & Härdtlein 2020,
p. 223). In the DIN ISO 14040, a product system is defined as a “collection of
unit processes with elementary and product flows, performing one or more defined
functions, and which models the life cycle of a product” (DIN ISO 14040, p. 11).
For this work, a product system is understood as the product itself, the required
inputs and outputs as well as the production and end of life treatment of the
product.

Modular product architecture

As a product can have many forms and serve many different purposes, the archi-
tecture can vary from simple forms with little functionality (e.g. a coin) to very
complex structures (e.g. a vehicle). The design of the product architecture is an
important aspect during the product development phase since the functionality
as well as the environmental and financial impacts of a product system highly
depend on the architecture of the product. Ulrich defines the product architec-
ture as “the arrangement of the functional elements, the mapping from functional
elements to physical components and the specification of the interfaces among
interacting physical components” (Ulrich 1995, p. 420). The functional elements
represent the parts of the product that define its functions and characteristics,
while the physical components are the actual parts that enable these functions
(Ulrich 1995, p. 420).
Müller (2013) describes the product architecture as the organization of how the
functional elements are brought together to form the final product. The smallest
element is defined as a part. Such a part is also described as the smallest physical
level of a product which cannot be disassembled any further without destroying
it. It usually shows no higher complexity and specialization for a certain func-
tion. Multiple parts form components, while multiple components form a module.
2.1 Modularity of Product Systems and Product System Fleets 11

Such a module is defined by its ability to realize one or multiple functions for
the product. (Müller 2013, p. 12) Kim and Moon (2019) describe the product
architecture similarly, as the final product is built out of multiple modules, while
each module is formed by multiple components. As the lowest level, Kim and
Moon (2019) do not include the level of parts but introduce a level for the utilized
material (see figure 2.1).

Figure 2.1 Visualization of product architecture (Kim & Moon 2019, p. 388)

For this work, the architecture of a product is based on the definition of Müller
(2013) and Kim and Moon (2019) and adapted to fit the definitions of a product
and a product system of the DIN ISO 14040. The term “product (system) struc-
ture” is used synonymously for a product’s (or a product system’s) architecture
in the following chapters. Figure 2.2 shows the structure of a modular product
unit and how it is embedded into a whole product system. In the example of
figure 2.2, the product unit consists of three different modules (A, B and C).
These modules represent the highest level of elements in the product architecture
below the level of the assembled product. Each module then is made from two
different components, which in turn also consist of two different parts. The prod-
uct units’ structure is formed from lowest to highest level: part  component
 module  product unit. Adjunct processes of the life cycle perspective, like
material extraction, production steps, the use stage and end of life treatment are
12 2 Theoretical Background and Technical Overview

associated to the product unit as well as input (materials, energy, etc.) and output
streams (waste etc.). All together, these elements of a product unit and supporting
processes form the entire product system.
As already defined in chapter one, a modular product design, as shown in
figure 2.2, consists of individual and independent modules that can be physically
detached from or combined with each other via defined interfaces (Gu & Sos-
ale 1999, p. 387 ff.). Wilhelm (1997) defines modularity as “a complex assembly
forming a closed function unit which permits specific differentiation and which,
as a consequence of defined interfaces (function, geometry), can be developed,
manufactured and assembled independently. Such an assembly must be interchange-
able and/or capable of alternative installation, it must represent an efficient unit in
terms of production and logistics, and it must require a minimum of modifications.”
(Wilhelm 1997, p. 147).
The more detailed the functions of the product are described, the more clearly
the individual functions the products modules need to provide can be under-
stood. The mapping from the required product functions to the modules can be
done in different ways. One option is the “one-to-one” approach, where each
function is provided by one module. Other approaches have multiple functions
combined within one module (“many-to-one”) or have many modules within the
product that all support one function (“one-to-many”). If a product consists of
a one-to-one mapping of modules and functions, it is called a modular prod-
uct architecture, while one-to-many or many-to-one indicate an integral product
architecture. (Ulrich 1995, p. 420 ff.) Interactions between the modules of a
product usually consist of the exchange of material flows, energy, or signal infor-
mation that lead to a higher functionality of the final product (Gu & Sosale 1999,
p. 387 ff.).
Jose and Tollenaere (2005) describe modularization as “an approach to orga-
nize complex designs and process operations more efficiently by decomposing
complex systems into simpler portions. It allows the designer to play with com-
binations of groups of components to develop and customize a larger quantity of
products.” (Jose & Tollenaere 2005, p. 371).

Product modularity including module alternatives

The complexity of modular products increases when for each module multiple
alternatives exist that provide the same function but differ in e.g. material choices
or shape. Thus, switching between module alternatives does not change the func-
tionality of the final product, but leads to different configuration options during
the product design. Figure 2.3 shows an exemplary product structure for a product
unit that is built from four different modules that each provide a unique function
 product system
architecture
level
product development
part part 1 part 2 part 3 part 4 part 1 part 2 part 3 part 4 part 1 part 2 part 3 part 4

component component 1 Component 2 component 1 component 2 component 1 component 2

module module A module B module C

life cycle perspecve

product unit

material extracon producon use phase end of life

2.1 Modularity of Product Systems and Product System Fleets

input streams output streams

(material, energy, …) (waste, …)

Figure 2.2 Definition of a modular product system structure

13
14 2 Theoretical Background and Technical Overview

to the product (module A, module B, module C and module D). Every module
offers multiple alternatives, where each alternative may lead to differences in its
environmental and/or financial performance compared to the initial module.
For module A, these alternatives are denoted as MA1 , MA2 , MA3 and MA4
(correspondingly for the remaining modules). Each module in turn consists of
three components (component A, component B and component C), while each
component has three alternatives (CA1 , CA2 and CA3 for component A). On
the lowest level of the product’s architecture, the level of parts, each component
consists of two parts (part A and part B), each having two alternatives (PA1 and
PA2 ). Regarding the entire product unit, 16 module alternatives are available on
the highest architecture level. Of these options four must be chosen to be part
of the final product, one for each necessary module. One level deeper into the
architecture, on the level of components, 144 alternatives are available of which
twelve must be selected for the product unit. On the lowest architecture level,
576 part alternatives form the basis of the product unit, here 24 parts need to be
selected in total.
To build an entire product, only one, but also always a minimum of one
alternative of the pool of alternatives, must be chosen. In this work, the terms
“alternatives”, “variants” and “options” are used synonymously to describe alter-
natives for modules, components, or parts with the same functionality but
different properties e.g. material or weight.
Each individual product unit of the possible product unit configurations is
further referred to as individual or unique “product configuration”, as each unit
is configured from one possible combination of the available module alternatives.
A higher modularity and a higher number of module alternatives lead to a higher
number of possible product configurations.

Degree of modularity and dependencies

In order to be able to assess the quantity of modules a product unit consists of
and determine the ideal degree of modularity, many studies have been performed.
Following Ulrich (1995), two modules are coupled if a change to one module
requires a change of the other module. It is also stated that two physically con-
nected components are almost always coupled. The physical connection between
two modules often leads to dependencies between the modules as the characteris-
tics of one module influence the features of the other module. For a product with
a fully modular product architecture, a change of one module does not affect the
other modules. Considering a more integral product design, a change of one mod-
ule can also lead to required changes to other modules. (Ulrich 1995, p. 422 ff.)
2.1 Modularity of Product Systems and Product System Fleets 15

modular product unit architecture

level
PA1 PA2 PB1 PB2
part A part B part

CA1 CA2 CA3 CB1 CB2 CB3 CC1 CC2 CC3

component A component B component C
component

MA1 MA2 MA3 MA4 MB1 MB2 MB3 MB4 MC1 MC2 MC3 MC4 MD1 MD2 MD3 MD4
module A module B module C module D
module

product unit

Figure 2.3 Modularity of product units (based on Jose & Tollenaere 2005, p. 374)

A high modularity and functional independence of the modules reduces changes

of the entire product, due to changes in selected functionality, to a minimum.
Ulrich (1995) distinguishes the performance of a product into local and global
performance characteristics. While the local performance of a function can be
optimized by enhancing the functionality of the module that is responsible for
this function (e.g. brighter lights for better sight), a global performance (e.g. the
fuel efficiency of a vehicle) depends on the characteristics of multiple modules.
Hence, the optimization of the global performance requires the consideration of
all modules and their interdependencies. (Ulrich 1995, p. 432 f.) Chen et al.
(1994) have shown that the independence between the individual modules of a
product can be increased by decreasing the interactions between these modules
(Chen, Rosen, Allen & Mistree 1994, p. 31).
The research of Gu and Sosale (1999) link a modular product architecture to
the goals of life cycle engineering (LCE). It is stated that to reach the objectives
of LCE with a modular product, “the relationships between the objectives and
the modules should be established” (Gu & Sosale 1999, p. 387). Following Gu
and Sosale (1999), a physical interaction between product modules is described
by the form of attachment, the positioning of the modules, the motion and the
containment. When individual components are grouped into modules, these inter-
actions need to be considered. For product design goals like high standardization,
mass customization and potential reuse and recyclability, Gu and Sosale (1999)
introduce a method for multiple criteria decisions to identify the ideal product
16 2 Theoretical Background and Technical Overview

modularity with minimal interactions between the modules. (Gu & Sosale 1999,
p. 387 f.)
Sosa, Eppinger and Rowles (2007) define modularity “as the level of inde-
pendence of a component from the other components within a product” (Sosa,
Eppinger & Rowles 2007, p. 1120). Hence, the modularity of a component
increases with a higher independence of other components, offering more degrees
of freedom. Figure 2.4 shows the hierarchical decomposition of a product into its
components.
By assessing the connections between individual components, the degree of
connectivity is quantified. Using the approach of graph theory, Sosa et al. (2007)
identify the most important components of the products network by evaluating
the centrality of the nodes. A high centrality of a node occurs when the respective
component is “directly connected” to many other components, “close to” all other
components and “between” many other components and therefore connecting
them. (Sosa et al. 2007)

Figure 2.4 Hierarchical decomposition of a product (Sosa et al. 2007, p. 1119)

Building on the design dependencies between components that are defined by

physical connections and flows of energy, material or information, Sosa et al.
(2007) also include virtual connections into their modularity assessment. Due to
these virtual connections, not only physical relations but also design dependen-
cies are considered. These design dependencies are represented by the distance to
other components and the connection the component enables between other com-
ponents. Three types of modularity are introduced that can be separately assessed
to measure the independence of modules within a product (Sosa et al. 2007):
2.1 Modularity of Product Systems and Product System Fleets 17

• Degree modularity:
Describes the physical connections between components. The more compo-
nents of a product are affected by a change of a component, the lower the
modularity of this component. If a change of a component affects no other
component, this component is “completely disconnected”, resulting in the
highest possible modularity.
• Distance modularity:
The distance between a specific component and other components also indi-
cates the level of modularity. The distance only depends on the direction of the
connection and not on the strength of the connection. A high distance modu-
larity indicates that a component is connected to other components by many
linking components making it an isolated component with a higher modularity.
• Bridge modularity:
The bridge modularity focusses “on those components that lie in the dependency
path of two components” (Sosa et al. 2007, p. 1122). If a component serves as a
bridge between many modules by connecting components, it becomes respon-
sible for the propagation of the design dependencies between the connected
components. The bridge modularity of a component is defined by its appear-
ances in the dependency paths of other components. The higher the number
of paths a component occurs in, the lower its bridge modularity (Figure 2.5).

Figure 2.5 Network of product components (Sosa et al. 2007, p. 1120)

By assessing these three different kinds of modularity, a deeper understand-

ing of the interdependencies between the modules of a product is achieved. This
18 2 Theoretical Background and Technical Overview

understanding helps optimizing the separation of components into modules. Fur-

thermore, the effect that changing one module has on the entire product can be
estimated. (Sosa et al. 2007)

Modular product systems

Based on the principles of modularity in products, it is also possible to define
modular product systems. It is required to transfer the adjunct and supportive pro-
cesses as shown in figure 2.2 into the structure of a product unit to expand the
modularity from product units to the entire product system. In this way, produc-
tion processes or material choices can also be modular and exchangable. Thereby
they contribute to the level of modularity of a product and expand the number
of possible product system configurations as well as the possibility of occurring
interdependencies between several choices of alternatives.

2.1.2 Multiple Product Systems and Population Fleets

The introduced product system perspective cannot only be applied to a single

product system, but also to multiple product systems. A collection of multiple
product systems is referred to as e.g. a product system population or a fleet of
product systems. Such fleets often come from the same product family, as e.g.
multiple vehicles of the same type form a fleet of vehicles.

Product families
Every single product unit within such a product family is described as a product
family member or product variant of the product family. The basis for the prod-
uct family, a modular product platform, is used to produce variations of a basic
product to create new variations of products by reconfiguring different module
alternatives. This approach is called “configurational product family design”, as
it is based on modular product architectures as new products are built from exist-
ing and standardized models. (Jiao, Simpson, & Siddique 2007) Simpson (2004)
defines a product family as “a group of related products that is derived from a prod-
uct platform to satisfy a variety of market niches” (Simpson 2004, p. 4). Simpson
(2004) further describes a product family as the result of a platform-based prod-
uct development where new product units of that product family are created by
“adding, substituting and/or removing” the modules of the basic product system
that provides the desired functionality (Simpson 2004, p. 5).
It is important to design the product structure in such a way that the resulting
modules lead to a high variety of possible product family alternatives to serve
2.1 Modularity of Product Systems and Product System Fleets 19

many different purposes from the existing modules to utilizing product families
in an efficient way. When designing product families and defining the module
partitioning, four major influences are described by Dahmus, Gonzalez-Zugasti &
Otto (2001): (1) market variance to cover different customer concerns, (2) market
variance to cover variety needs after the purchase, (3) technology change for
updating the product design and (4) design for X to consider life cycle criteria.
Yang et al. (2014) describes one central aspect of product families as the
improvement of commonality by defining fixed modules that are shared by all
product units of the product family. These modules form the basis while further
modules help to create different functionalities by varying additional modules.
(Yang, Yu & Jiang 2014)

Advantages of product families

Building product fleets based on product families from a shared platform leads to
many advantages for the OEM of these products, as many different authors have
shown. Most of these advantages occur due to the utilization of standardized
modules across a large scale of product systems.
One advantage is an easier and more cost-efficient mass customization of
product units to fit customer needs. Pandremenos et al. (2009) describes the mod-
ularity in production as the basis for mass customization of vehicles. A modular
production allows the OEM to “pre-combine a large number of components into
modules” while still allowing individual combinations of modules (Pandremenos,
Paralikas, Salonitis & Chryssolouris 2009, p. 148). For OEMs, e.g. in the auto-
motive sector, the utilization of product fleets based on product families also
saves expenses when building many product units on the same platform with
a high commonality of components. Such product families have many shared
parts which simplifies production and makes it more cost-efficient. (Dahmus,
Gonzalez-Zugasti & Otto 2001) These reduced production costs are also realized
due to components or modules not only being similar, but sharing many produc-
tion processes. Also, with modular product architectures, product derivates can
be developed quicker and more cost-efficiently as developers focus on special-
ized modules to customize and differentiate the new product from the product
family. (Jose & Tollenaere 2005, p. 371 ff.) Simpson (2004) also adds that not
only development time is reduced, but also the time and effort that is required
for testing and certification of new product derivates. Modular product families
also lead to a better ability of upgrading product units. (Simpson 2004, p. 4) In
the extensive literature review, Bataglin and Ferreira (2020) also identified a pos-
itive impact of product families on the reusability of product units as well as the
20 2 Theoretical Background and Technical Overview

reparability and quality issues as broken modules can more easily be replaced.
This supports the longevity of a product unit. (Bataglin & Ferreira 2020)
Kim and Moon (2019) describe a better recyclability as an advantage of prod-
uct families with shared modules, as the disassembly and recycling processes can
be optimized and standardized to improve the product recovery. (Kim & Moon
2019) Additionally, Yang et al. (2014) describes that modular product fleets and
their shared modules with a high commonality use their economy of scale to bring
environmental and financial optimization into a large amount of product units in
a short time. Wedler and Vietor (2019) point out that with a modular concept
for autonomous vehicles (e.g. separation in life and drive modules), more pas-
senger transport scenarios can be realized without having to increase the number
of vehicles (Wedler & Vietor 2019).

Fleets of product systems

As described by various authors, product families based on a modular product
platform with variable modules offer many advantages. However, besides these
positive aspects, more complexity is also added when modularization is applied
for mass customization. When not only a single product unit is considered but a
fleet of multiple units, the optimization process needs to cover all units and not
only one single unit of the product fleet. As every unit of the product fleet is
configured freely and individually from all available alternatives per module, the
total number of possible options for a fleet configuration grows quickly compared
to the options for only a single product unit, as shown in equation 1.1. Figure 2.6
shows the structure of a product system as introduced in figure 2.2 expanded
as a fleet of multiple product systems and expanded with a modular product
architecture as introduced in figure 2.3.

2.2 Life Cycle Thinking

Product systems, as introduced in the previous chapter, are described and assessed
in different ways and from different perspectives. In order to fully consider all
environmental and financial aspects that are related to the product system, it is
important to consider the entire life cycle and not only parts of it. Following this
life cycle perspective, the environmental life cycle assessment and the financial
life cycle costing can be performed.
 fleet of product systems 1
…
n
product development
2.2 Life Cycle Thinking

life cycle perspective

product unit

material extraction production use phase end of life

input streams output streams

(material, energy, …) (waste, …)

n = number of product unit within fleet

Figure 2.6 Fleet of multiple product systems

21
22 2 Theoretical Background and Technical Overview

Life cycle perspective

Product life cycle concepts are descriptive models that depict the economic and
ecologic effects of the individual life cycle stages of a product or a product sys-
tem (Mateika 2005, p. 8). Today, there are many ways to describe the life cycle
of a product, focusing on different stages or performance indicators. The basic
model of a product’s life cycle has been introduced by Hofstätter (1977), which
describes financial figures like the revenue of a product over its market time.
Since then, other life cycle concepts have been developed that also cover addi-
tional aspects of the life cycle. These additional aspects include the production
and end of life stage of a product as these stages also define the properties and
performance of the product over its lifetime. (Herrmann 2010, p. 70 ff.) Forma-
tive life cycle concepts that go beyond the concept of Hofstätter (1977) are e.g.
the integrated life cycle concept by Pfeiffer and Bischof (1981), which includes
the period before the market entry, or the systemic product life cycle by Klenter
(1995) that expands to a production and an end of life stage (Mateika 2005, p. 9
ff.).
A different form of life cycle modelling is found in flow-oriented concepts.
These models focus on the material and energy flows from the beginning to the
end of a life cycle in a linear model or even in a circle back to the beginning.
Such a circular model applies when recycling is used to retrieve secondary mate-
rial from the old product at the end of a life cycle. (Herrmann 2010, p. 65)
A flow-oriented life cycle concept can roughly be divided into the phases of
material extraction, production and manufacture, use and service and finally dis-
posal and recycling (Keoleian, Kar, Manion & Bulkley 1997, p. 5). In figure 2.7,
this typical flow-oriented life cycle model is depicted. The consecutive steps of
“production stage”, “use stage” and “end of life stage” and their characteristic
contents describe the life cycle of a product from cradle to grave.
The typical life cycle, as depicted in figure 2.7, presents the boundary frame-
work for the environmental and financial assessment of a product system in this
work. A detailed introduction to the different forms of life cycle concepts is found
in Hermann (2010).

2.2.1 Environmental Life Cycle Assessment

Assessing the life cycle of a product system regarding its ecological performance
is done by analyzing the inputs and outputs of a product system and linking
them to environmental impacts. As already stated in section 1.1, it is important
to evaluate the environmental performance of a product system along the entire
2.2 Life Cycle Thinking 23

cradle to grave

production phase use phase end of life phase

extraction of raw materials production of operating materials reutilisation and recycling
& & &
manufacturing of the utilisation or operation of the disposal of the waste- and scrap-
(intermediate) product product materials

Figure 2.7 Flow-oriented life cycle model (based on Broch 2017 and Keoleian et al. 1997)

life cycle to avoid shifting environmental burdens between life cycle stages. The
global standard for evaluating the ecological life cycle performance are life cycle
assessments (LCA) based on the DIN ISO norms 14040 and 14044. Such an LCA
is distinguished in three different types. The first type is a gate-to-gate analysis
linked to the processes related to the activities of a single company or factory.
The second type is a more detailed analysis of a single process and the third type
the life cycle assessment of an entire product system from cradle to grave. This
third type covers all processes linked to this product, going beyond the borders
of a single company or factory. (Hermann 2010, p. 151 f.)
A life cycle assessment is a scientific method to quantify the impacts of a
product system along the entire life cycle. It is a suitable tool to compare envi-
ronmental impacts of two different alternatives. LCAs also help to avoid problem
shifting between life cycle stages and provide a detailed database to support deci-
sion making regarding the product system, e.g. on how to reduce environmental
impacts (Koffler 2007, p. 12 and UNEP 1996, p. 5 ff.).
In the DIN ISO norm 14040, an LCA is defined as follows:

[A life cycle assessment is the] “compilation and evaluation of the inputs, outputs
and the potential environmental impacts of a product system throughout its life cycle.”
(DIN ISO 14040, p. 7)

The general approach of an LCA is based on an iterative framework consisting

of four steps: goal and scope definition, inventory analysis, impact assessment
and interpretation of results (DIN ISO 14040, p. 4). After the LCA has been
interpreted, a report for the intended audience has to be written. This report shall
address the scope, data, assumptions and limitations of the study to make the
calculated results and their interpretation as transparent as possible. (DIN ISO
14040, p. 32) Figure 2.8 depicts this life cycle assessment framework and its
24 2 Theoretical Background and Technical Overview

iterative steps. This norm-based approach of an LCA is referred to as a “regu-

lar” or “conventional” LCA. The functional unit needs to be the same for each
comparative assessment to make sure that the result of an LCA is comparable to
other LCA studies. This functional unit defines the reference unit to which all the
inputs and outputs are related to by describing the functions the analyzed product
shall fulfill. The DIN ISO norm 14040 provides an example of a specific number
of dried hands serving as the functional unit in order to compare paper towels
and blow dryer systems. (DIN ISO 14040, p. 23 f.)

life cycle assessment framework

step 1 step 2 step 3

goal and scope
inventory analysis impact assessment
definition

interpretation

step 4

Figure 2.8 Life cycle assessment framework (based on DIN ISO 14040, p. 16)

Step 1: Goal and scope definition

In the step of goal and scope definition, the intended application and the rea-
son why the LCA is performed have to be explained. The addressed audience
needs to be named as well as the motivation behind the LCA (e.g. comparing
two alternatives or identifying ecological hotspots of a single product). Also, the
functional unit, the system borders (e.g. cradle-to-gate or cradle-to-grave) and
the intention on whether the study is supposed to be published or not have to
be defined. (DIN ISO 14040, Kloepffer & Grahl 2009 and Koffler 2007, p. 13)
For the scope definition, details regarding the procedure of the LCA have to be
described. Among them are e.g. the applied allocation procedures, cut-offs, the
selected impact categories, the methodology of impact assessment, the assump-
tions and limitation of the study and the quality requirements for the input data.
(DIN ISO 14040, p. 23)
2.2 Life Cycle Thinking 25

Step 2: Inventory analysis

The step inventory analysis of the life cycle is defined as follows:

“Inventory analysis involves data collection and calculation procedures to quantify

relevant inputs and outputs of a product system. The process of conducting an inventory
analysis is iterative. As data are collected and more is learned about the system, new
data requirements or limitations may be identified that require a change in the data
collection procedures so that the goals of the study will still be met.” (DIN ISO 14040,
p. 25 f.)

Figure 2.9 shows the flow of materials as well as the inputs and outputs of a
production process with three consecutive processes. The dotted line represents
the system boundaries for this inventory analysis.

energy, water, air & process materials

inputs

raw
materials product
process A process B process C

outputs

wastes, heat & emissions

Figure 2.9 Processes with input and output flows (based on Broch 2017, p. 15 and Kloepf-
fer & Grahl 2009, p. 67)

For the inventory analysis it is crucial that the processes and material flows
of the product system are correctly analyzed and modelled. Those processes and
flows are then matched with the identified inputs and outputs like energy, raw
materials or emissions and wastes. The goal of the inventory analysis is the calcu-
lation of the elementary flows, which represent the energy or materials that either
come from or go into the environment without any further treatment. The outcome
is the added results of the elementary flows per functional unit. Also, the cut-off
procedures need to be documented to reduce complexity without compromising
the relevant influences. (Kaltschmitt & Schebek 2015, p. 215 ff.)
26 2 Theoretical Background and Technical Overview

Step 3: Life cycle impact assessment

The third step of an LCA transfers the findings of the inventory analysis to the
calculation of the potential environmental impact of each identified and quantified
reference flow of the product system (Koffler 2007, p. 16). The impact assessment
consists of three mandatory (selection, classification, characterization) and three
optional elements (normalization, grouping, weighting) as depicted in figure 2.10.
During the selection step, the relevant impact categories, category indicators and
characterization models are chosen. The classification serves as the link of the
results from the inventory analysis to an impact assessment. Due to this step,
it is possible to determine whether an elementary flow has a specific environ-
mental impact. The classification associates the results of the inventory analysis
with their midpoint environmental impact (see figure 2.10). (DIN ISO 14040 and
Koffler 2007)

mandatory elements

selection classification characterization

selection of impact categories,
calculation of category indicator
category indicators & assignment of LCI results
results
characterization models

+
optional elements
(normalization, grouping, weighting)

Figure 2.10 Life cycle impact assessment (based on DIN ISO 14040, p. 30)

The different midpoint impact categories are expressed by means of a specific

category indicator. This indicator characterizes all the elementary flows that are
assigned to the same impact category into the same unit of measurement. The
category of global warming is e.g. expressed by the indicator kg of CO2 eq. (DIN
ISO 14044 and Koffler 2007, p. 18 f.). As each elementary flow has an individ-
ual impact on the selected midpoint category, the results of the inventory analysis
are converted—using a characterization factor—into the dimension of the cate-
gory indicator. Regarding the midpoint impact category of climate change, every
elementary flow of the inventory analysis potentially has an individual effect on
global warming. The emission of one kg methane for example has the same
global warming potential (GWP) as 25 kg of CO2 , giving methane a CO2 equiv-
alent of 25. (Kaltschmitt & Schebek 2015, p. 276) Figure 2.11 visualizes the
2.2 Life Cycle Thinking 27

process from the inventory analysis, classification and characterization towards

the category indicator for the example of climate change.

inventory analysis classification characterization category Indicator

CO2

CH4 quantification of the global warming

midpoint category of
increase of infrared potential in [kg CO2
climate change
N2O radiation eq.]
…

Figure 2.11 Classification and characterization of the impact assessment (based on Broch
2017, p. 17)

Table 2.1 lists selected environmental midpoint impact categories and their
respective indicators and reference units. The optional steps of normalization,
grouping and weighting are applied to further increase the comparability of the
LCA results. The normalization focuses e.g. on regions, grouping sorts categories
according to similar impacts and weighting is applied to rate different categories
and add them into one total indicator. (Koffler 2007, p. 20 f.) Further information
on impact categories can be found in Herrmann 2010, p. 157 ff.

Table 2.1 Selected environmental impact categories, indicators, and reference units (based
on Stranddorf, Hoffmann & Schmidt 2005, p. 36)
impact category impact indicator reference unit
climate change global warming potential kilogram carbon dioxide
(GWP) equivalents [kg CO2 eq.]
stratospheric ozone ozone depletion potential kilogram trichlorofluoromethane
depletion (ODP) equivalents [kg CFC-11 eq.]
summer smog formation photochemical ozone kilogram ethylene equivalents [kg
formation (PCOF) C2 H4 eq.]
acidification acidification potential kilogram sulfur dioxide
(AP) equivalents [kg SO2 eq.]

Step 4: Interpretation
The fourth step of an LCA is the interpretation of the results, which is supposed
to assist the decision making regarding the analyzed product system. Based on the
findings, alternatives or replacements can be assessed and the most ecologically
28 2 Theoretical Background and Technical Overview

friendly configuration is identified. Furthermore, it is important to evaluate the

LCA regarding completeness, sensitivity of critical parameters and consistency.
(DIN ISO 14040, p. 22)
Life cycle assessments for complex product systems are conducted by using
specialized databases and software for the process modelling, inventory analysis
and impact assessment. Examples and overviews of suited software and databases
are found in Broch (2017), Kaltschmitt & Schebek (2015) and Kloepffer & Grahl
(2009).

2.2.2 Life Cycle Costing

In addition to the environmental life cycle assessment, the life cycle of a product
system including its financial performance has to be evaluated in order to be able
to evaluate its eco-efficiency and/or eco-effectiveness. Having assessed a measure
for the CO2 reduction of a product system regarding its environmental impact
(by means of an LCA) and its financial impact, that measure can be ranked with
regard to its CO2 reduction potential per spent budget. The economical evaluation
is based on the same life cycle definition as the LCA in order to make sure that all
relevant financial flows are covered and the comparability with the LCA results
is sufficient (Rebitzer & Hunkeler 2003).
In the field of cost accounting, there are many different approaches to evaluate
the costs of a product. The concept of life cycle costing (LCC) is applied to
assess a product regarding the economical expenses over the entire life cycle.
The term of life cycle costing is used for many different forms of accounting
the preliminary costs, the costs for the operation and the subsequent costs of a
product in a holistic manner. Analyzing the costs of a product along its entire
life cycle helps to assign the incurring costs and revenues to the different life
cycle stages. (Hoch, Heupel & Kachel 2016, p. 331) A strict definition of LCC
(compared to the DIN ISO norms 14040 and 14044 for LCA) does not exist.
However, it is scientific consensus that the subsequent costs should be integrated
and trade-offs between the life cycle stages due to cost substitution have to be
considered (Ulmschneider 2004, p. 50). Rebitzer and Hunkeler (2003) define life
cycle costing as follows:

[Life cycle costing is the] “assessment of all costs associated with the life cycle of a
product that are directly covered by any one or more of the actors in the product life
cycle (supplier, producer, user/consumer, EOL-actor), with complimentary inclusion
2.2 Life Cycle Thinking 29

of externalities that are anticipated to be internalized in the decision-relevant future.”

(Rebitzer & Hunkeler 2003, p. 254 f.)

Following Rebitzer and Hunkeler (2003), costs are understood as expenses within
the operational service process of all goods within a defined period consisting of
basic and imputed costs (Paul 2015, p. 204).
Following Herrmann (2010), the evaluation of the costs along the entire life
cycle of technical products is especially relevant if the share of the total costs
within the use stage and/or end of life stage is relatively high compared to the
acquisition costs. If the life cycle costing only contains the costs of the different
life cycle stages, life cycle costing in the narrow sense is performed. The scope
can be expanded if the revenues are also included in the calculations. (Herrmann
2010, p. 131 f.)
Life cycle costing can be accounted from different points of view, which leads
to different results if certain costs or revenues are either included or excluded.
Two typical perspectives for an LCC are the manufacturer’s point of view and
the customer’s point of view, each having a different focus on the product’s cash
flow. If the target is to reduce the total cost of ownership (TCO) of a product,
those two different cost perspectives lead to different approaches. This is due to
the different financial hotspots or cost types that those perspectives have and that
result in different concepts of cost optimization. (Ulmschneider 2004, p. 49 ff.)
Evaluating the total cost of ownership of a product is not only relevant for
customers that strive to identify the most cost-efficient alternative of different
interchangeable products, but also for the manufacturers that produce these goods.
Being able to offer a product with lower costs during the use stage, this techno-
logical advantage will be reflected in the selling price. A cost-reduced use stage
is one possible way for companies to compensate detriments such as higher wage
costs when producing in developed countries. (Cerri et al. 2014) Kloepffer and
Ciroth (2011) state that “environmentally preferable products often have higher
purchasing costs, whereas the LCC may be much lower (examples: energy saving
light bulbs, low energy houses, and cars)” (Kloepffer & Ciroth 2011, p. 99). The
aspect of covering multiple life cycle stages makes LCC a suitable tool for cus-
tomers to compare different products. Given the customer behavior of comparing
costs over the entire life cycle, the manufacturers should be interested in reducing
the TCO of a product as far as possible as otherwise it will be less successful on
the market compared to alternative products (Kloepffer & Ciroth 2011, p. 99).
The results of life cycle costing can either be calculated in retrospective or in
advance for a future product. Each result is influenced by many uncertainties
30 2 Theoretical Background and Technical Overview

that increase with a higher complexity of the product and with a longer lifetime
(Scope, Ilg, Muench & Guenther 2016).
Figure 2.12 shows a representation of life cycle costs by the German associa-
tion of engineers (Verein Deutscher Ingenieure, VDI). In line with the definition
of a product’s life cycle, the life cycle costs result in the addition of the individual
costs of each life cycle stage. Following the definition of the VDI, the life cycle
costs are calculated by equation 2.1:

Li f ecyclecosts = Costs Pr oduction + CostsU se + Costs Endo f Li f e (2.1)

The costs of the different life cycle stages each consist of various cost types
like development and manufacturing costs (production stage) or operating and
maintenance costs (use stage). As figure 2.12 depicts, these costs are paid by
different cost owners. While the costs of the production stage are paid by the
manufacturer, the costs of the use stage are paid by the operator or customer.
The shift of the financial responsibility is marked by the transfer of the ownership
of the product from the manufacturer to the customer including a profit for the
manufacturer. (VDI 2005, p. 5)

Figure 2.12 Life cycle costing by the VDI (VDI 2005, p. 5)

2.2 Life Cycle Thinking 31

Costs that occur along the life cycle of a product are either classified as fixed
or variable costs. Fixed costs are expenses that do not change when an influencing
variable is altered while the variable costs are affected by such a change. (Plinke,
Rese & Utzig 2015, p. 29) An example of variable costs are the costs for the
input material that is needed to produce goods. The demand for the product then
denotes the influencing variable. As a result of a rising demand and a constant
price for a fixed unit of the input material, the total expenses for input material
will increase (as shown in figure 2.13).
total costs [€]

variable

total costs
costs
costs
fixed

product demand [quantity]

Figure 2.13 Fixed and variable costs (based on Plinke et al. 2015, p. 30)

Another form of cost categorization is the differentiation into direct costs and
overheads. While fixed and variable costs are separated by their reaction to an
influencing parameter, direct costs and overheads can be distinguished by cause.
Direct costs are costs that are directly assigned to one produced unit of goods
while the overheads cannot be linked to the explicit production of just one prod-
uct. As overheads are caused by more than one unit of goods, they are usually
evenly distributed among all produced goods. (Plinke et al. 2015, p. 36)

Costs of the production stage

The costs of a product that occur during its production stage are split into the
sub-stages of product development, production, and distribution. Adding up these
cost categories leads to the prime costs of the product. Together with the profit
margin, these costs constitute the acquisition costs for the customer. Usually, the
development costs are overheads and not explicitly linked to the development of
one product unit. The production costs are defined by the direct material costs,
32 2 Theoretical Background and Technical Overview

acquisition costs of vendor parts and a material overhead as well as the manu-
facturing direct costs and the manufacturing overhead. While the material costs
include the expenses for raw material, the material overhead includes expenses
for inhouse logistics. The manufacturing costs consist of the personnel costs and
costs for e.g. energy, electricity and maintenance that are required to keep the
production running. Those costs typically have a share of direct costs and over-
head costs. The costs for distribution account for e.g. warehousing or transport.
(Nickenig 2016; Rudorfer 2017; Schlink 2014) The costs of the production stage
are summarized based on Nickenig (2016), Rudorfer (2017) and Schlink (2014)
with equation 2.2, with CPP being the costs of the production stage, CDE the
development costs, CPR the production costs and CDI the distribution costs

CP P = CDE + CP R + CDI (2.2)

Costs of the use stage

The costs that incur during the use stage of the product consist of the operating
costs, maintenance and service costs and possible additional taxes, contractual
costs and insurance expenses. The operating costs are defined by the costs for the
operating materials as well as lubricants or other additional consumable materi-
als. The maintenance and service costs include e.g. spare parts, repair costs and
service costs for regular inspections. (Herrmann 2010, p. 134) The costs of the
use stage are summarized based on Herrmann (2010) and VDI (2005) with equa-
tion 2.3, where CUP represents the costs of the use stage, COP the operating costs,
CMS the costs for maintenance & service and CTI the taxes and insurance costs.

CU P = C O P + C M S + C T I (2.3)

Costs of the end of life stage

The costs of the end of life stage (CEP ) of a product usually consist of the
incurring costs for dismantling (CDM ), recycling (CRE ) and disposal (CDI ) (see
equation 2.4) (Herrmann 2010, p. 134; VDI 2005).

CE P = CDM + CRE + CDP (2.4)

Figure 2.14 depicts the life cycle stages of a product system (a vehicle in this
case) and the corresponding costs that occur in each stage for the example of a
vehicle as a product system.
2.3 Sustainable Development and Corresponding Legislation 33

• operating costs • contractual costs

customer
costs

• maintenance costs • tax costs

• service costs • insurance costs
life cycle
product

production use phase recycling

manufacturer

• development costs • production costs • dismantling costs

costs

• material costs • personnel costs • recycling costs

• vendor part costs • distribution costs • disposal costs

Figure 2.14 Life cycle costing—stages and costs (based on Herrmann 2010, p. 134)

2.3 Sustainable Development and Corresponding

Legislation

Due to a growing need for energy and resources, humans influence their envi-
ronment (see section 1.1). Production and consumption of products and services
cause emissions and waste streams which affect the environment and have an
impact on the local and global eco systems. Environmental targets and regula-
tions are derived to avoid or reduce the negative impacts on the environment. The
following subchapters describe relevant environmental impacts, the need for sus-
tainable development and corresponding legislation for environmental protection.
A special focus lies on the regulation for the automotive industry to comply with
global environmental targets.

2.3.1 Environmental Impacts and Impact Assessment

The earth’s environment is a complex system of atmosphere, hydrosphere and

lithosphere and the relations of those with all living and non-living organisms
(Günther 2018). It is impacted by the extensive use of natural resources and
energy through human activities (Herrmann 2010, p. 13). The growing popula-
tion and the growing demand for energy and resources induce a rapid growth
of the economic output, but also a growing impact of those actions on the envi-
ronment and its natural resources (Kaltschmitt & Schebek 2015, p. 1 f.). These
34 2 Theoretical Background and Technical Overview

natural resources are defined as renewable and non-renewable resources, phys-

ical space, environmental media such as air, water, soil, flowing resources like
geothermal energy, wind, waves, solar radiation, and biodiversity. The utilization
of these resources occurs either due to extraction of these resources (environ-
mental source) or due to absorption of emissions and wastes into these resources
(environmental sink), in both ways leading to an environmental impact. (Umwelt-
bundesamt 2016) The anthropogenic impact on the environment is understood as
measurable changes in the ecosphere, which are considered either positive or
negative from the human perspective. These environmental impacts comprise
technical, economical, socio-cultural, and ecological changes. (Kaltschmitt &
Schebek 2015, p. 15 f.) The chain of effects from the cause of the impacts to the
final responses consists of the drivers, pressures, states, impacts and responses of
an environmental impact. Figure 2.15 depicts this chain of effects.
The drivers are the fundamental cause for the chain of effects. Typical exam-
ples for such drivers are the growing global population or the growing demand
for energy. The following environmental pressures are the actual and potential
interferences with the natural environment that are triggered by the drivers and
are distinguished between indirect (e.g. emissions from burning fuels) and direct
(e.g. deforestation) impacts.

drivers pressures states impacts responses

• growth of global • deforestation • global average • melting of glaciers • cleaning of exhaust
population • emissions of temperature • extinction of gases
• demand for energy substances • number of species species • renewable energy

Figure 2.15 Chain of effects for environmental impacts (based on Smeets & Weterings
1999, p. 6 ff.)

The states of the environment then describe, based on scientific parameters,

the condition of the natural environment for a fixed point in time. Possible key
figures that express the environmental state are the global average temperature
or the number of species. The impacts (either short-term or long-term) describe
the actual effects of the environmental pressures, such as the melting of glaciers
or the extinction of species. The last link in the chain of effects for environmen-
tal impacts are the responses, which interact with all preceding steps, leading to
interactions. This phase describes the human reaction to the previous environmen-
tal impacts and includes (if the impact is considered negative) e.g. cleaning of
2.3 Sustainable Development and Corresponding Legislation 35

exhaust gases or the utilization of renewable energy sources. (Smeets & Weterings
1999, p. 6 ff.)
While the human influences on the environment (the pressures) can exactly
be quantified, the resulting consequences (the environmental impacts) cannot.
These impacts are usually complex and often influence the environment on a
global scale—in long-term perspective and in multiple environmental categories.
Hence it is complicated to measure, quantify, and predict these impacts precisely.
(Kaltschmitt & Schebek 2015, p. 4 f.)
The life cycle impact assessment (LCIA) midpoint-damage framework of the
UNEP/SETAC life cycle initiative shows that the human caused environmen-
tal pressures lead to environmental midpoint impacts. These midpoint impacts
e.g. include human and eco-toxicity, eutrophication, land use, (a-)biotic resource
depletion, climate change, acidification, or ozone depletion. These effects in
return affect e.g. the human health, the (a-)biotic resources, the (a-)biotic natural
environment and the (a-)biotic man-made environment, as figure 2.16 depicts.
(Jolliet et al. 2004, p. 395)

inventory results midpoint endpoint area of protection

climate change
stratospheric ozone human health
depletion
human toxicity
particulate matter
formation
elementary flows

photochemical ozone natural

formation environment
ecotoxicity
acidification
eutrophication
land use
water use natural
resources
abiotic resource use

Figure 2.16 Life cycle impact assessment midpoint framework (Hauschild & Huijbregts
2015, p. 9)
36 2 Theoretical Background and Technical Overview

An important question when dealing with environmental impacts is whether

the impacts can be categorized as a positive impact on the environment or a neg-
ative impact that causes an environmental problem. An environmental impact
constitutes an environmental problem if e.g. the time period, spatial effect,
urgency and irreversibility of the damages that occur change the environment
compared to the previous state in an undesired way judging from the human
perspective (Kaltschmitt & Schebek 2015, p. 17).
The GHG emissions are known to be a major factor for the anthropogenic
climate change. Their primary effect is the radiation absorption by the molecules
in the atmosphere and are indicated as carbon dioxide equivalents (CO2 eq.)
(Kloepffer & Grahl 2009, p. 195 ff. and 224). Accounting for roughly 76 %
of the global anthropogenic carbon dioxide equivalents, carbon dioxide is the
most important greenhouse gas. Consequently, carbon dioxide has the biggest
impact on the anthropogenic climate change. It is followed by methane (16 %)
and nitrous oxide (6.2 %) (IPCC 2014, p. 6). For Germany in 2018, approx.
78.7 % of the methane emissions and 81.6 % of the nitrous oxides emissions are
emitted by the agricultural sector and waste disposal, while 93.2 % of the carbon
dioxide emissions are energy related emissions coming from industry processes
(see figure 2.17) (Umweltbundesamt 2019a). One reason for the high emission
level of carbon dioxides is the high share of fossil-based energy carriers like coal,
oil, and gas with 81.0 % (in 2014) for the provision of the world’s primary energy
demand (IEA 2016, p. 64).

2.3.2 Sustainability and Sustainable Development

Environmental targets can be classified within the concept of sustainable devel-

opment. The term “sustainable development” was defined in the Brundtland
Report “Our Common Future” by the World Commission on Environment and
Development (WCED) in 1987 as follows:

“Sustainable development is development that meets the needs of the present without
compromising the ability of future generations to meet their own needs.” (WCED
1987, p. 41)

At the United Nations Conference on Environment and Development in 1992

in Rio de Janeiro, the United Nations (UN) published the “Rio Declaration on
Environment and Development” wherein it is stated that “Human beings are at
the centre of concerns for sustainable development. They are entitled to a healthy
2.3 Sustainable Development and Corresponding Legislation 37

Share of agriculture in
Germany: 81.6%

Share of agriculture
and waste disposal in
6% 2%
Germany: 78.7%
Share of energy in
16% Germany : 93.2%

76%

Carbon dioxide Methane Nitrous oxid F-Gases Other

Figure 2.17 Shares of greenhouse gases and main causes (based on IPCC 2014, p. 6 &
Umweltbundesamt 2019a)

and productive life in harmony with nature.” (United Nations 1992, p. 1) and “the
right to development must be fulfilled so as to equitably meet developmental and
environmental needs of present and future generations.” (United Nations 1992,
p. 2). This underlines the concept of the Brundtland definition.
In 1998, the Enquete Commission of the German parliament published their
final report on the concept of sustainability for the protection of humans and the
environment defining the concept of the “triple bottom line” as the simultaneous
integration of ecological, economic, and social goals for a successful sustainable
development (Deutscher Bundestag 1998). Only if all three dimensions are suffi-
ciently fulfilled at the same time, sustainable development can be achieved (see
figure 2.18).
One of the biggest initiatives regarding the persecution of sustainable devel-
opment are the “Sustainable Development Goals” of the UN. In 2015, the UN
introduced 17 different goals for a sustainable development of the Earth’s soci-
ety until 2030, aiming at more secure and prosper living conditions. With goals
number 12 (“responsible production and consumption”) and number 13 (“cli-
mate action”), two goals are directly targeted at environmental protection and a
sustainable product development. (United Nations 2015b)
38 2 Theoretical Background and Technical Overview

Figure 2.18 Depiction of

the triple bottom line (based
on Elkington 2018) People
Social development

Sustainability

Planet Profit
Environmental Economic
development development

Eco-efficiency and eco-effectiveness

To derive solutions with an optimized effectiveness, the term “effectiveness”
needs to be distinguished from the term “efficiency” as both terms describe the
success of a measure regarding a desired target. “Efficiency” brings the dimen-
sions of input and output into relation and defines the productiveness of the
input. “Effectiveness” on the other hand describes the degree of target attainment
by expressing the ratio of the actual state and the desired state. (Eichhorn &
Merk 2016, p. 183; Hauschild 2015) While efficiency is a relative expression,
effectiveness is an absolute expression of e.g. improvement measures.
The term “eco-efficiency” has been coined by the World Business Council for
Sustainable Development (WBCSD) during the “Earth Summit” of 1992 in Rio
de Janeiro, Brazil, and is defined as follows:

“Eco-efficiency is achieved by the delivery of competitively priced goods and services

that satisfy human needs and bring quality of life, while progressively reducing eco-
logical impacts and resource intensity throughout the life-cycle to a level at least in
line with the Earth’s estimated carrying capacity.” (WBCSD 1995, p. 4)

Eco-efficiency is often defined as the ratio of economic value and environmental

impact (based on Schaltegger & Sturm 1990, p. 281 ff.):

Economicvalue
EcoE f f iciency = (2.5)
Envir onmentalimpact
2.3 Sustainable Development and Corresponding Legislation 39

However, building on the definition of the term “productivity”, which expresses

the ratio of output and input, the definition of equation 2.5 must be put into the
perspective of efficiency as a measure for improvement. Therefore, to understand
whether the eco-efficiency of a product is sufficiently high, it must be compared
to the eco-efficiency of a reference product as defined in the following equation:

EcoE f f iciencyactual
Evaluating EcoE f f iciency = (2.6)
EcoE f f iciencyr e f er ence

To improve the eco-efficiency, either the economic value must be increased for
the same environmental impact or the environmental impact for a constant eco-
nomic value needs to be reduced (see equation 2.5). Measures for environmental
protection become relevant for profit-oriented companies in two different ways.
Either if it adds a value to the company’s business and increases its produc-
tiveness or if self-set or legislative environmental limits must be met. Since this
shall be done with the highest cost-efficiency, the eco-efficiency plays a key role
for companies that follow environmental goals. (Schrack 2016, p. 34 f.) The term
“eco-efficiency” is used in the context of a relative improvement of sustainability.
In comparison to eco-efficiency, the term “eco-effectiveness” is defined by
Jakobsen (1999) as:

“[…] the total impact on environment when the consumers need or demand is satisfied
by alternative fulfilment of the function in question. One way of fulfilling a function is
said to be more eco-effective than by fulfilling it in a different way, if it gives a larger
contribution to sustainability of the eco-system in question than the second way of
fulfilling the function.” (Jakobsen 1999)

Additionally, Braungart, McDonough and Bollinger (2007) state that “the concept
of eco-effectiveness proposes the transformation of products and their associated
material flows such that they form a supportive relationship with ecological systems
and future economic growth. The goal is not to minimize the cradle-to-grave flow
of materials, but to generate cyclical, cradle-to-cradle “metabolisms” that enable
materials to maintain their status as resources and accumulate intelligence over time
(upcycling).” (Braungart, McDonough & Bollinger 2007). Figure 2.19 visualizes
the differences between the terms eco-efficiency and eco-effectiveness. While
eco-efficiency strives to reduce the negative impact for the ecological system,
eco-effectiveness expresses the positive impact.
While the eco-efficiency improves the relative environmental impact of a prod-
uct, it is not guaranteed that a product with an increased eco-efficiency is a
sustainable product. The “rebound effect” causes a situation where the efficiency
40 2 Theoretical Background and Technical Overview

+
Impact to ecological system

Eco-effectiveness
optimizing positive impact

time
Eco-efficiency
reducing negative impact
-

Figure 2.19 Eco-efficiency vs. eco-effectiveness (based on Braungart, McDonough &

Bollinger 2007; Koeijer, Wever & Henseler 2017)

gains are compensated by a more extensive use of the improved product which
equalizes its relative improvements. In such cases, eco-efficiency does not con-
tribute enough to sustainable development. And even without the rebound effect,
an increased eco-efficiency does not necessarily yield a sustainable product as the
product can still have an environmental impact which is too high to be sustained
over time. (Hauschild 2015) While the eco-efficiency only expresses the relative
improvement of the environmental impact of a product, the eco-effectiveness of
a product describes its ability to contribute to an absolute sustainability, consid-
ering a possible rebound effect due to a higher utilization or a growing consumer
group. (Hauschild 2015; Hauschild, Herrmann & Kara 2017; Kara, Hauschild,
Herrmann 2018)
An approach to quantify the absolute sustainability and to define a “safe oper-
ating space for humanity” is described by the concept of “planetary boundaries”
by Johan Rockström et al. (2009). A total of nine different environmental cat-
egories (e.g. freshwater use or biosphere integrity) are analyzed and individual
limits for a safe global operating space defined. Surpassing the limits of these
safe operating spaces implies that the Earth has left the controllable state for
human activities. According to Rockström et al. (2009), this safe operating space
has already been exceeded for the category of climate change. (Rockström et al.
2009)
2.3 Sustainable Development and Corresponding Legislation 41

While the definition of sustainability by the triple bottom line (see figure 2.18)
brings the three aspects of “people”, “Earth” and “profit” together, eco-efficiency
only describes the relation between “Earth” and “profit”. Thus, the aspect of
social development is not included in this perspective. In contrast to sustainability,
eco-efficiency only focuses on the economic value, the factor of social sustain-
ability is not quantified and not considered in the concept. Another difference
between sustainability and eco-efficiency is that for sustainability, all three dimen-
sions shall be optimized or increased. For eco-efficiency, the selected dimensions
“Earth” and “profit” are related by determining the ratio of both aspects. The
triple bottom line on the other hand does not define such a qualitative relation
between its dimensions.

2.3.3 General Environmental Legislation

Environmental targets are set when an environmental problem is detected and

has to be diminished or if the path of a sustainable development is followed. An
environmental target is defined as a limit for an environmental burden (e.g. emis-
sion reduction targets) or a vision for an environmental state that is strived for.
Such environmental targets are the foundation for legislative activities which are
a compromise between environmental goals and socio-economic targets. Envi-
ronmental goals can either be defined for specific elements (e.g. limitation of
CO2 emissions) or spatial regions, but can also be expressed more generally (e.g.
limiting the rise of the global mean temperature). (Feess 2018) One possible
way to reach such an environmental goal is the production and consumption of
eco-effective products and services.
Figure 2.20 shows an example of a possible definition of an environmental
target. Following the concept of Rockström et al. (2009) and starting with the
goal of preserving the safe operating space of the Earth, it is important to limit
climate change to a maximum of 2° C. This leads to the need of reducing GHG
emissions and finally results in a limitation of CO2 emissions.
After environmental targets have been defined, general guidelines and prin-
ciples follow in order to set effective regulations. The Enquete Commission of
the German parliament defined five guiding rules for material flows to comply
with the given goals for a sustainable development. These five rules can be seen
as guidelines on how to reach a sustainable development: (Deutscher Bundestag
1998, p. 25)
42 2 Theoretical Background and Technical Overview

preservation of reducing
Earth`s safe greenhouse gas cutting CO2
operating space limiting climate emissions emissions to a
change to max. minimum
Defining an 2 °C environmental target

Figure 2.20 Defining an environmental target

• The degradation rate of renewable resources shall not be higher than their
recovery rate,
• Non-renewable resources shall only be utilized in a way that either an equiv-
alent form of renewable resources is created or the productivity of (non-)
renewable resources is increased,
• Material flows into the environment shall be scaled to the absorption capacity
of the environment,
• Material flows into the environment shall only occur in a suitable period for
the environment to react,
• Threats and unacceptable risks for human health by anthropogenic causes shall
be avoided.

Damages already caused to the environment shall be removed following the “pol-
luter pays principle”. This principle implies that the environmental damages shall
be removed by those organizations that are responsible for causing them. Fol-
lowing the “precautionary principle”, generally known and proven hazards to the
environment shall be avoided in advance, just as well as potential threats to the
environment (Herrmann 2010, p. 49 f.). For economically oriented companies,
the “efficiency strategy” is the most suitable approach to reduce environmental
burdens. The efficiency strategy assumes that natural resources and the envi-
ronment are scarce input factors whose consumption needs to be reduced over
time (Schrack 2016, p. 2). Measures that help to reach environmental targets e.g.
include an increased energy efficiency of products or factories, applying renew-
able energy forms or energy carriers with a lower environmental impact and the
utilization of secondary materials coming from recycling streams (IPCC 2014).
Many legislative approaches to operationalize environmental targets are imple-
mented in national or multi-national laws. A prominent example of a global-wide
cooperation for climate protection is the Paris Agreement of 2015. In this agree-
ment, 195 signatories have agreed to “holding the increase in the global average
2.3 Sustainable Development and Corresponding Legislation 43

temperature to well below 2° C above pre-industrial levels and pursuing efforts to

limit the temperature increase to 1.5° C above pre-industrial levels” (United Nations
2015a, p. 3). In the EU, the “seventh environment action program” describes
the ambition to reduce the greenhouse gas emissions by 20 % and to increase
the share of renewable energies as well as the energy efficiency by 20 % by
2020 (European Parliament 2013, p. 2). Until 2030, the targets are tightened to a
reduction of at least 55 % regarding GHG emissions compared to 1990. (Euro-
pean Commission 2020c) This target is also implemented in the European Union
Emissions Trading System (EU ETS), which regulates the GHG emissions of sev-
eral industries since 2005 following the “cap and trade” approach. Within the EU
ETS, a limited amount of emission certificates is made available for each partic-
ipating company, which in turn can trade these emission rights on a marketplace.
The EU ETS is decreasing the allowed GHG emissions in general by reducing
the amount of emission certificates that enter the market every year. This leads
to a situation where the emissions are reduced by those companies that achieve
them in the most cost-efficient way (European Commission 2015). In 2019, the
European Commission introduced their plans for a “European Green Deal” that
sets out the vision for a decarbonized Europe by 2050. An intermediate target for
2030 is defined including a reduction of GHG emissions by 55 % compared to
1990. (European Commission 2019)
On a national level, the German constitution e.g. proclaims in article 20a that
the state of Germany must protect the natural environment and the basis of all
life to meet the responsibility it has for its current and future inhabitants (MoJ
2020d). The German “Bundes-Klimaschutzgesetz” from 2019 sets the reduction
of GHG emissions to at least 65 % by 2030 and therefore follows the targets of
the European Commission. The German climate action plan 2050 sets the target
to meet the Paris Agreement by reducing the GHG emissions by up to 95 % by
2050. (FME 2016)

2.3.4 Environmental Regulations for the Automotive

Industry

The methodology of this thesis is developed as an example of an application in the

automotive industry. This industry is chosen as the transport sector is responsible
for a high share of the total CO2 emissions as shown in section 2.3. In Germany,
approx. 17 % of the GHG emissions are caused by transports (Umweltbundesamt
2015). Due to the high environmental impact of the automotive industry, many
aspects of a vehicle are subjected to regulations. These regulations aim to control,
44 2 Theoretical Background and Technical Overview

limit, and reduce the emissions of the automotive industry over time. In this sub-
chapter, an overview of the worldwide regulations regarding the emissions from
a vehicle is given. Along the major stages of the life cycle of an automobile (pro-
duction stage, use stage and end of life stage), different environmental impacts
occur that lead to regulations for each life cycle stage.

Regulations for the production stage

Regarding the production of a vehicle or its sub-parts, no laws exist that define
limitations of environmental pollutants that directly address the automotive indus-
try. Instead, the factories and production sites of the OEM or their suppliers are
bound to the general environmental laws and regulations of the region they are
located at. For example, in Germany such regulations include:

• Abwasserabgabengesetz (AbwAG): Regulation regarding the discharging of

waste waters into waters (MoJ 2020a)
• Bundes-Immissionsschutzgesetz (BimSchG): Federal act for pollution control
regarding harmful effects on the environment caused by air pollution, noise,
vibration, and similar phenomena (MoJ 2020b)
• Chemikaliengesetz (ChemG): Federal act on the protection against hazardous
substances, defining which materials are classified as hazardous and how these
materials must be treated (MoJ 2020c)

Factories in Germany always have to comply with those kinds of environmen-

tal regulations to maintain a secure production of e.g. a paint shop or a vehicle
assembly. However, these regulations are subject to all production sites in Ger-
many and do not include specific targets for the automotive industry. On top of
these regulations, German OEMs have set up individual campaigns to further
reduce their environmental impact by reducing e.g. the amount of fresh water
used, the primary energy demand, or the amount of waste per produced vehicle.
(VDA 2014) As an example, the Volkswagen AG targets to reduce the environ-
mental impact of its vehicle production by 45 % by 2025 compared to 2010. This
reduction involves greenhouse gas emissions, use of fresh water, waste production
and energy demand. (Volkswagen AG 2017)

Regulations for the use stage

As the use stage of a vehicle causes emissions that impact the environment as
well as the human health, numerous regulations exist that control the emissions
of a vehicle during its operation. The legislation in this field is divided into two
2.3 Sustainable Development and Corresponding Legislation 45

different focus areas: (1) regulation of air quality by limiting air pollutants in
exhaust gases and (2) regulation of greenhouse gases emitted by vehicles.
The legislation regarding air quality limits the emission of air pollutants per
driven kilometer. These limits must not be exceeded by any vehicle. The Euro-
pean Union for example restricts the emission of seven different components
of exhaust gases. The following table shows an example of the current limita-
tion values for those substances for diesel and gasoline fueled vehicles under the
newest Euro Standard 6. (European Parliament 2007) (Table 2.2)

Table 2.2 Limitation values for air pollutants set by Euro Standard 6 (European Parliament
2007)
Substance Gasoline Diesel
Carbon monoxide (CO) 1000 mg
km 500 mg
km
Total hydrocarbons (HC) 100 mg
km –
Non-methane hydrocarbons (NMHC) 68 mg
km –
Oxides of nitrogen (NOx ) 60 mg
km 80 mg
km
Combined mass of total hydrocarbons and oxides of – 170 mg
km
nitrogen (HC + NOx )
mg mg
Mass of particular matter (PM) 4.5 km 4.5 km
Number of particular matter (PN) 6 ∗ 1011 number
km 6 ∗ 1011 number
km

As these substances do not have a relevant impact on climate change, the

regulation of air pollutants is not further considered in this thesis.

Regulations for the end of life stage

The end of life legislation is different for each region and market around the
world. In Europe for example, the directive 2000/53/EG of the European Parlia-
ment regulates the end of life treatment of vehicles. Its goal is to harmonize the
country specific processes of the EU members to increase the reuse and recy-
cling rate of returning vehicles. This directive determines the framing conditions
regarding (1) the prevention of waste, (2) the collection of end of life vehicles, (3)
the treatment of end of life vehicles, (4) the targets for reuse and recovery quotes,
and (5) the coding standards for dismantling information. Since 1st January 2015,
at least 95 % of the vehicle (regarding mass) must be reused or recycled. Further-
more, a bigger focus on design for recycling is demanded. (European Parliament
2000)
46 2 Theoretical Background and Technical Overview

Regulations regarding greenhouse gas emissions from vehicles

In order to make the automotive industry contribute to the decarbonization, many
countries and regions have added stricter regulations regarding the emission of
GHG. While the target of reducing GHG emissions is the same for every region,
the approach and the target values are different. As shown in figure 2.21, over
three quarters of all CO2 emissions are caused by the energy, transport and man-
ufacturing industries. In the European Union, the transport sector is responsible
for almost 30 % of those emissions. 72 % of them are caused by road-based
transportation. Figure 2.21 also shows how the transportation-related emissions
are distributed among the different transportation modes. Most of those emissions
are caused by passenger cars. Together with the light duty vehicles, the passenger
cars cause 72.6 % of the road-based transport emissions. (European Parliament
2019)
In the largest automotive markets (European Union, United States of America
[USA], and China) the greenhouse gas legislation for vehicles is divided into
(1) regulations for passenger cars plus light duty vehicles and (2) regulations for
heavy duty vehicles. The regulations between those classes differ regarding the
limit values, time frames and calculation methods.
The following sections give an overview of the greenhouse gas legislation for
passenger cars in the three named markets.

Within this group

• Passenger cars: 60.7%
1.0% • Light duty vehicles: 11.9%
13.4%
• Heavy duty vehicles: 26.2%
• Motorcycles: 1.2%

13.6%

72.0%

Road transportaon Water navigaon Civil aviaon Other

Figure 2.21 Transport-related CO2 emissions in the European Union (based on: European
Parliament 2020)
2.3 Sustainable Development and Corresponding Legislation 47

Greenhouse gas legislation in the European Union

In contrast to the regulation of air pollutants, the regulation of greenhouse gas
emissions does not apply to every single vehicle. Instead, for each OEM, a fleet
wide average is calculated and set as a limit value. This limit value is defined
as grams of CO2 per driven kilometer in the tailpipe emissions. The regulations
443/2009 and 333/2014 of the European Parliament limit the greenhouse gas
emissions to 95 gCkmO2 by 2021, which applies as the allowed average emission
level of an OEM’s fleet. (European Parliament 2009, European Parliament 2014)
l
Converted to fuel consumption, this limit value corresponds to approx. 4.1 100km
l
of gasoline or 3.6 100km of diesel. In concession to different types and classes
of vehicles, the average weight of an OEM’s fleet is taken into account for the
calculation of the emission limitation. Compared to the overall average of vehicles
in the market, an OEM can have a higher emission level if the vehicles are heavier
than the average. An additional weight of 100 kg leads to an allowed excess of
the emission limit by 3.33 gCkmO2 . . (Transportpolicy 2020a)
These values are measured for each of the OEM’s vehicles by test cycles
to guarantee comparable results. In Europe, the New European Driving Cycle
(NEDC) and the Worldwide Harmonized Light Vehicles Test Procedure (WLTP)
are used. While the current values are measured by the NEDC, the WLTP will
replace the NEDC over time. (European Parliament 2009, European Parliament
2014) Both test cycles consist of several sequences of defined velocities over a
defined time. While the NEDC is 1,180 seconds long, the WLTP lasts for 1,800
seconds. During these 1,800 seconds, the NEDC covers 10.97 km and the WLTP
23.27 km. The WLTP has a higher mean and maximum velocity with 47 km/h to
34 km/h and 131 km/h to 120 km/h. (Volkswagen AG 2020) Figure 2.22 shows
a comparison of the NEDC and WLTP regarding the velocity over time.
If an OEM reduces the CO2 emissions by introducing an “eco-innovation”,
credits regarding the emission limit are applied, even if the eco-innovation is not
effective in the test cycle. Such credits can sum up to 7 gCkmO2 for each OEM.
(Transportpolicy 2020a) This incentive is introduced to increase the innovation
regarding emission reduction technologies. To further increase the introduction of
more environmentally friendly vehicles, “super credits” for zero or low emission
vehicles (ZLEV) are accounted with a maximum of 7.5 gCkmO2 for each OEM. A
ZLEV is defined as a vehicle with an emission level of less than 50 gCkmO2 . . In
2020, a ZLEV is counted as two vehicles, in 2021 as 1.67 vehicles and in 2022
as 1.33 vehicles. (European Commission 2020a) If the defined emission limit
(95 gCkmO2 under consideration of weight impact and eco-innovations) is exceeded
48 2 Theoretical Background and Technical Overview

Figure 2.22 Comparison of NEDC and WLTP (Volkswagen AG 2020)

by an OEM, a penalty has to be paid. This penalty amounts to e95 per exceeded
gram of CO2 and per sold vehicle. (Transportpolicy 2020a)
In the regulation 2019/631 by the European Parliament, the emission goals
for the time after 2020 are defined. For 2025, a further reduction of the emission
limit by 15 % is defined, for 2030 a reduction by 37.5 %. (European Parliament
2019) For ZLEVs, a new credit system is applied, starting in 2025. This credit
system is activated when the share of ZLEVs of all newly registered vehicles of
an OEM in the given year is above a threshold value. In 2025, this value is at
15 %, from 2030 on at 35 %. For every percentage point an OEM exceeds those
threshold values, the specific CO2 emission target will also be increased by one
percent with a maximum of up to 5 %. An even greater weighting applies when
ZLEVs are introduced in markets with a slow growth in the ZLEV segment.
(European Commission 2020b)
In a further outlook, the European Commission will collect and evaluate data
on real driving emissions from cars using the on-board fuel consumption mon-
itoring devices. If required, further steps shall be defined to adjust the OEM’s
emission targets. The expansion of the emission reporting from tailpipe emissions
2.3 Sustainable Development and Corresponding Legislation 49

to full life cycle reporting is also discussed. By 2023, the European Commis-
sion is supposed to assess a methodology that is used to collect and report data
regarding the CO2 emissions along the entire life cycle of a vehicle. (European
Commission 2020b) This is a first step into the direction of setting regulations
not only to the emissions of exhaust gases but to the whole environmental foot-
print of a vehicle product lifetime. Further information about the CO2 standards
for new vehicles in Europe is found in ICCT 2017.

Greenhouse gas legislation in China

In China, the emissions of greenhouse gases are not regulated by the emissions
per driven kilometer but by the fuel consumption efficiency of a vehicle. The
emission standards are defined by the ministry of industry and information tech-
l
nology. Since 2016, phase IV is active, which sets the standard to 5 100km
for 2020. This level of fuel efficiency is applied to vehicles with a total mass
below 3,500 kg. In total, 16 different classes of vehicle weight are defined which
all have individual efficiency targets. Corresponding to the European legisla-
tion, this is done to consider the impact of different vehicle classes. While in
Europe the emission limit is based on a fleet average, the weight-oriented fuel
efficiency in China must be met by each vehicle individually. A fleet average
in China is defined in a metric called “Corporate Average Fuel Consumption”
(CAFC). OEMs must comply with both the individual mass-oriented limit and
l
the CAFC-oriented efficiency, which is currently set to 5 100km , which converts
gC O2
to 117 km . . The basis for the fuel efficiency calculation is the NEDC test
cycle. (Transportpolicy 2020b)
Like the European Union, China also has implemented incentives for innova-
tive powertrain technologies and eco-innovations. Vehicles with an electric range
of more 50 km are accounted as five vehicles into the CAFC. Vehicles with a
l
combined fuel efficiency of below 2.8 100km are counted as three vehicles. Eco-
l
efficiencies with an off-cycle effect are credited to a maximum of 0.5 100km . . It
is also possible for OEMs to transfer a surplus of fuel efficiency from the present
into a following year if the current emission level is below the specific limit.
(Transportpolicy 2020b) For 2025, a further decrease to a fuel efficiency of only
l
4 100km or 93 gCkmO2 is enacted. (ICCT 2020)

Greenhouse gas legislation in the United States of America

The regulation of greenhouse gas emissions of vehicles is the responsibility of
the Environmental Protection Agency (EPA) and the US department of trans-
portation. Beyond the nationwide rules, the state of California has its own limits
50 2 Theoretical Background and Technical Overview

based on the regulation of the California Air Resources Board (CARB). The
limits are defined as emission limitation and fuel efficiency standards and are
applied to vehicles with a mass below 4,536 kg. They are measured by the Fed-
eral Test Procedure 75 (FTP-75) and weighted with the highway cycle. In contrast
to the legislation in Europe and China, the US regulations do not take the vehi-
cle weight into account, but instead vary the detailed emission limits with regard
to vehicle size. By calculating the vehicle’s footprint, individual limits are set.
(Transportpolicy 2020c)
For 2025, a target value of 163 gC O2
mile is defined. This is a reduction of the
mi
previous limit by 35 % and corresponds to a fuel efficiency of 54.5 gallon or
l
4.3 100km or to an emission level of 107 gCkmO2 . Since the US limit standard
is designed for passenger vehicles and light duty trucks, a single limitation for
passenger cars results in a limit of 91 gCkmO2 in 2025 when transferred to the
NEDC. (Transportpolicy 2020c) In the state of California, the greenhouse gas
emission limits are lower than for the rest of the USA with limits of 183 gC O2
mile
gC O2
in 2020 and 144 mile in 2025 (Transportpolicy 2020d).
Like Europe and China, the USA also has a credit system for eco-innovations
that are effective off-cycle:

“Emission reduction compliance credits include air-conditioning system technology,

flexible fuel vehicle deployment, off-cycle technologies, incentives for electric vehicles,
and “game-changing” technologies installed on pickup trucks. Of these credits, only
the air-conditioning credits and some off-cycle technology credits reflect real-world
emission reductions that are not included on the compliance test cycles. The others
reduce the overall stringency of the standards, without corresponding reductions in
real-world emissions.” (Transportpolicy 2020c)

Figure 2.23 shows a summary of worldwide CO2 emission legislations for vehi-
cles, normalized to the NEDC driving cycle of the European Union. In this figure,
the emission regulations of California are equivalent to the level in China. It
becomes visible that the European Union has, as of today, set the strictest emis-
sion regulations. The European Union is also the region that has set regulations
for the longest time span into the future with targets for 2030. China and USA
have set limit values only up to 2025.
2.4 Life Cycle Engineering and Mathematical Optimization 51

Figure 2.23 Comparison of CO2 emission limitations worldwide (ICCT 2020)

2.4 Life Cycle Engineering and Mathematical

Optimization

The tools of life cycle engineering and mathematical optimization can be used
to help developers and decision makers to design sustainable and eco-efficient
product systems and product system fleets. In this way, it is possible to trans-
fer the requirements of environmental targets of legislations and regulations into
the development of new product systems. The following sections explain the
fundamentals of LCE and mathematical optimization.

2.4.1 Life Cycle Engineering

In order to be able to optimize a product system in terms of its performance

along the entire life cycle, the product needs to be designed in a life cycle ori-
ented scope already during its development phase. LCE is an established concept
to incorporate environmental targets into the product development phase. Being
aware of the entire life cycle of a product and its defining phases during the
process of decision making is an essential part for a holistic reduction of envi-
ronmental impacts. This approach is also called “life cycle thinking”. (Koffler
52 2 Theoretical Background and Technical Overview

2007, p. 6) Due to the product design paradox (see figure 2.24), it becomes obvi-
ous that the degree of freedom for changes to the product system is the highest
during the early stages of the product development.

100%
Product Knowledge

Modification Cost

Freedom of action

0%
Timescale

Figure 2.24 The product design paradox (O’Reilly et al. 2016, p. 751)

As more time during the product development passes, this degree of freedom
decreases, while the costs for changes to the product system increase. This leads
to the situation that decisions that influence important parts or functionalities of
a product need to be made as early as possible. Unfortunately, this opposes the
knowledge about the product system, as the product knowledge increases over
time. Important and influencing decisions for the product system often need to
be made with relatively little product information, enhancing the importance of
the life cycle perspective and LCE in general. As a consequence, the developed
methodology in this study is meant to help decision makers during the early
phases of product planning and product development to make decisions that lead
to reduced environmental burdens over life cycles, when the costs for changes
are still relatively low.
The concept of Life Cycle Management (LCM) can be selected to apply this
principle to a product system. In the background report for the United Nations
Environment Programme (UNEP) guide to Life Cycle Management in 2006,
LCM is defined as follows:
2.4 Life Cycle Engineering and Mathematical Optimization 53

„LCM is the application of life cycle thinking to modern business practice, with the
aim to manage the total life cycle of an organization’s products and services towards
more sustainable consumption and production. LCM is about systematic integration of
product sustainability e.g. in company strategy and planning, product design and devel-
opment, purchasing decisions and communication programs.” (Jensen & Remmen
2006, p. 10)

Along the concept of LCM, the principle of LCE is a framework for many
approaches that cover the field of the optimization of eco-efficiency or eco-
effectiveness. LCE comprises engineering approaches that are meant to oper-
ationalize the life cycle perspective in a way that decision support for the
development phase of a product can be provided (Alting 1995). LCE is applied
to develop products that have a reduced environmental impact by minimized pol-
lution and waste. Evaluation methods for the environmental impact are required
as well as the integration of the life cycle perspective and engineering activities
into the stage of product development to achieve this goal (Kaluza et al. 2017).
Hauschild, Herrmann and Kara (2017) introduced an “Integrated Framework
for Life Cycle Engineering” that integrates product development and manu-
facturing, meaning the earliest stages of a product’s life cycle, in the context
of the planetary boundaries and the concept of absolute sustainability and
eco-effectiveness. This framework (shown in figure 2.25) is positioned in the
dimensions of temporal concern and environmental concern.
Based on a top-down approach, LCE is described with a scope of
multiple products and their entire life cycle stages. All main aspects of
these stages (development, raw material extraction, production, after-sales,
reuse/remanufacturing/recycling) are covered. (Hauschild, Herrmann & Kara
2017) In view of this new framework, LCE is defined as:

“[…] sustainability-oriented product development activities within the scope of one to

several product life cycles. The methods and tools used in LCE must support reducing
the total environmental impact associated with technology change and volume increase
from one product generation to another, in order to ensure that new product technolo-
gies stay within their environmental space as derived from the planetary boundaries.”
(Hauschild, Herrmann & Kara 2017, p. 6)

A well-known approach to quantify the environmental impact and to put the focus
on the human made technology factor is the IPAT equation. This equation was
introduced by Ehrlich and Holdren (1971) and describes an approach to measure
the environmental impacts based on the global population, the human affluence,
and the technology (see equation 2.7). The main objective of the IPAT equation is
 54

Figure 2.25 Life cycle engineering framework (Hauschild, Herrmann & Kara 2017)
2 Theoretical Background and Technical Overview
2.4 Life Cycle Engineering and Mathematical Optimization 55

the deconstruction of the individual factors that result in the final environmental
impact. By understanding which factors are drivers of environmental impacts and
how these factors are interacting, strategies can be developed to mitigate the
impacts. The IPAT equation shows that the environmental impact (I) increases if
one of the three defining factors is raised. These factors are the population factor
(P), the affluence factor (A) and the technology factor (T ). As all three factors are
in a direct relation, an achieved decrease in one of these factors is compromised
by an increase of another factor. Hence, a growing population can equalize the
efficiency improvements made within the technology factor.

I = P∗ A∗T (2.7)

As economic companies do not have influence on the global population or the

affluence, their only focus is to reduce the technology impact. (Ehrlich & Hol-
dren 1971) Ideally, the reduction of this factor is high enough to compensate
gains in the other factors to comply with global environmental targets and to
stay within the limits of absolute sustainability (meaning eco-effective prod-
ucts) (Kara, Hauschild, Herrmann 2018). The efficiency is derived by using the
inverse value of the technology factor: T1 . . Regarding the environmental con-
cern, the concept of LCE influences the technology factor of the IPAT equation.
(Hauschild, Herrmann & Kara 2017)
The Kaya identity, which was introduced by Kaya and Yokobori (1997), fur-
ther deconstructs the IPAT equation. The Kaya identity states that the global CO2
emissions are the result of the factors population, gross domestic product (GDP)
per capita and energy consumption. Again, the reduction of population or GDP
per capita are not desired targets, requiring the reduction of either the energy
consumption or the CO2 emissions per energy consumption or energy provision.
The Kaya identity is expressed in equation 2.8, with F being the man-made
CO2 emissions, G the GDP per capita, P the population factor and E the energy
consumption. (Kaya & Yokobori 1997)

G E F
F=P∗ ∗ ∗ (2.8)
P G E

Figure 2.26 shows the concept of Hauschild, Kara and Røpke (2020) on how
absolute sustainability is integrated into life cycle engineering based on the IPAT
equation. In order to be able to not only improve a single product regarding its
environmental impact (relative eco-efficiency) but to consider absolute limits, it
is required to also include a growing market for the product by an increase of the
56 2 Theoretical Background and Technical Overview

factors population and/or affluence. When the total environmental impact of an

entire product generation remains within the limits of the planetary boundaries by
not exceeding its available emission budget, it contributes to eco-effectiveness.

Figure 2.26 Eco-effectiveness and life cycle engineering (Hauschild, Kara & Røpke 2020,
p. 3)

The IPAT equation and the Kaya identity point out that in order to reduce GHG
emissions, the efficiency of products has to be improved or the consumed energy
source freed from GHG emissions, or, ideally, both. Those reductions need to
outperform the remaining factors to lower the overall environmental impact and to
contribute to eco-efficiency. Transferred to the economic world and technological
industries, there is a huge need for environmentally optimized products that reach
absolute limits that are ambitious enough to support absolute sustainability. Thus,
newly developed products need to contribute to sustainability in absolute terms
and not only be more sustainable than the product they aim to replace.
2.4 Life Cycle Engineering and Mathematical Optimization 57

2.4.2 Optimization Approaches for Decision Support

Optimization approaches from the field of operations research are applied to

support decision makers with regard to many aspects of product systems concern-
ing their financial and environmental aspects. The following section introduces
the general principle of optimization methods and problem complexity that are
necessary to select a suitable approach for the optimization of the relevant
parameter.

Operations research and optimization

The term “operations research” is defined as a discipline that addresses the anal-
ysis of complex problem statements within the scope of a planning process to
support the decision-making process by using mathematical methods. The typi-
cal process of operations research follows the steps of (1) problem analysis, (2)
evaluation of targets and options for action, (3) mathematical modelling, (4), data
gathering, (5) problem solving and (6) evaluation of solution. These steps are
iterative loops that are constantly revised during the entire process. In a narrow
sense, operations research is often defined as the mathematical modelling of deci-
sion support problems and the development of algorithms to identify solutions for
the designed models. (Domschke, Drexl, Klein & Scholl 2015 p. 1 f.)
In a mathematical sense, an optimization problem is solved by identifying
the optimal solution of all possible options. All solutions must be comparable
regarding certain criteria to be able to classify the quality of each solution and to
identify the ideal solution. The comparability of solutions is achieved by defining
a specific objective that has to be reached. (Ellinger, Beuermann & Leisten 2003,
p. 2 f) The optimal decision-making process is described as follows:

“While considering the decision-restrictions and impacts, identify the decision that
fulfills the specific objective the most.” (Papageorgiou, Leibold & Buss 2015, p. 1)

Operations research offers the required tools to enable a systematic decision

support. It is fundamental to formulate a precise problem definition using math-
ematical notifications for the objective functions and secondary restrictions in
order to be able to use these tools. It is important to differentiate between given
input parameters and the sought parameter that is to be optimized with the objec-
tive function. This objective function (or target function) is usually expressed with
either a minimization or maximization function of the target value. (Papageorgiou
et al. 2015, p. 1 f.) Equation 2.9 shows a general formulation of the instance of an
optimization problem over R with an optimization of ϕ over F. In this equation,
58 2 Theoretical Background and Technical Overview

ϕ is the objective function and F the feasible region where every x ∈ F is per-
mitted. When a maximization of the objective function is desired, equation 2.10
is added to the objective function, in the opposite case, a desired minimization
of the objective function, equation 2.11 is applied. (Gritzmann 2013, p. 1 f)

n ∈ N ∧ F ⊂ G ⊂ Rn ∧ ϕ : G → R ∧ opt ∈ {min, max} (2.9)

 
x ∈ F → ϕ x ∗ ≥ ϕ(x) (2.10)

 
x ∈ F → ϕ x ∗ ≤ ϕ(x) (2.11)

Such optimization problems are solved with optimization algorithms or heuristics.

An algorithm is defined as follows:

“An algorithm is an exactly defined processing description to solve a given problem

or a special group of problems. Typically, an algorithm is characterized by a limited
amount of actions that are applied consecutively and are repeated in a determined
way.” (Kastner & Schildt 2005, p. 83)

Heuristics are described as general approaches for the search of an ideal solution
that accept intermediate deteriorations of the solution to leave a local optimum in
search for a better solution. However, a heuristic will not result in a guaranteed
optimum, as it is not possible to determine how far off the selected solution is
from the overall optimum. (Suhl & Mellouli 2013, p. 13)
The selection of either an algorithm or a heuristic depends on the type and
size of the given problem. In the field of operations research, many different
types of algorithms and heuristics exist that can be adapted and applied to a given
problem statement. Among the most spread procedures for solving optimization
problems are linear optimization (e.g. simplex approach), nonlinear optimiza-
tion (e.g. Lagrange multiplier), dynamic optimization and integer programming
(e.g. branch and bound). Popular (meta-)heuristics are e.g. tabu search, greedy
algorithm, or genetic algorithms. (Zimmermann 2008)

Complexity and problem size

The combinatorial problem size for a modular product system increases quickly
given many module options, as already shown in section 2.1. The large num-
ber of possible combinations requires the selection of an appropriate solution
approach. The number of possible combinations represents the search space or
2.4 Life Cycle Engineering and Mathematical Optimization 59

solution space of the optimization problem. Next to the type of the problem state-
ment, the size of the search space influences the complexity class of the problem.
The computational complexity theory (CCT) deals with the question if a given
problem that is described by an algorithm (and thus can be calculated) also can
be practically solved with limited computational capacity. CCT also strives to
identify the most efficient algorithm to solve the given problem.
Figure 2.27 shows that the practically solvable problem types form a subgroup
of the calculable problem types which in turn form a subgroup of all problem
types. Limiting factors for problem types to be practically solvable are time and
calculation capacity and depend on the size of the search space and the required
calculation operations. (Ernst, Schmidt & Beneken 2016) The complexity of a
problem and the behavior of an algorithm is described by the “Big O Notation”.
Due to the big O notation, it becomes possible to classify the performance of
an algorithm depending on the required computation time and space to solve the
given problem. (Chivers & Sleightholme 2018)
The defining feature of the big O notation of a problem statement is always
the biggest mathematical term of a function. Forthefunction f (x) = x 3 +10x 2 +
 2  results in f (x) = O x . . (Rubinstein-Salzedo 2018)
11x+17, the big O notation 3

A complexity
 of O x is significantly quicker to solve than a complexity of
O x 3 . Still, basically all polynomial problem statements are practically solvable
with an existing algorithm. For exponentially growing problems or algorithm
types on the other hand, the required calculation time is increasing rapidly. (Ernst
et al. 2016) Table 2.3 gives an overview of different big O notations and their
complexity.

Figure 2.27
Categorization of problem all problem
types (based on Ernst et al. types
2016)

calculable praccally solvable

problem types problem types

An algorithm for the complexity class of O(n) is programmed in such a way

that each of the n elements of the input data need to be processed only once
or a constant number of times. Doubling the input data hence leads to a twice
60 2 Theoretical Background and Technical Overview

 
as long calculation time. For a complexity class of O n 2 , , every element has
to be processed twice (e.g. sorting algorithm). Doubling the input data leads
to a processing duration of four times the original duration. For a data set N
and a complexity O(n), , a computer that calculates a thousand times faster
can process an input data that is thousand
  times bigger (1000N ) at the same
time. For a complexity class of O n 2 , , a thousand-fold calculation capacity
leads to an increase of only 32 times the original input data (32N ) and for
O(2n ) only to N + 10. The quickly increasing calculation time and the little
effect of a higher calculation speed of exponential complexity classes mark a
crucial difference between exponential and polynomial complexity classes. While
polynomial complexity classes are usually practically solvable, the exponential
classes are usually not. (Ernst et al. 2016)

Table 2.3 Complexity

Big O notation Complexity
classes and big O notation
(based on Chivers & O(1) Constant complexity
Sleightholme 2018) O(n) Linear complexity
O(log n) Logarithmic complexity
 
O n2 Quadratic complexity
 
O n3 Cubic complexity
O(x n ) Exponential complexity
O(n!) Factorial complexity

This situation is the foundation of the categorization of problems into the

classes of P and NP. Problems that fall into the category P are all decision prob-
lems that can efficiently be calculated within a polynomial runtime. It is solvable
if there is an algorithm with a complexity of O( p n ), where p n is a polynomial
function of an arbitrary degree. Hence, all problems that cannot be solved within
a polynomial time are sorted into the category NP (nondeterministic polynomial
time). (Ernst et al. 2016) If the ideal solution for a decision support problem
is desired, the problem type has to be in the problem class P and a suitable
algorithm with a polynomial runtime has to be applied.

2.5 Conclusions Regarding the Theoretical Background

Summarizing the findings of this chapter, the extensive use of carbon-based

energy sources to cover the high energy demand of human activities causes an
2.5 Conclusions Regarding the Theoretical Background 61

enormous impact on the GHG emissions and thereby on the change of the global
mean temperature. Legislations on global, continental, and national level have
been installed to control the GHG emissions to reach sustainable development
and absolute sustainability. This includes strict regulations for the automotive
sector, where GHG emissions need to be drastically reduced to comply with
future regulations.
In order to improve product development and to design products that are in
line with these decarbonized restrictions, the tools of life cycle engineering need
to be applied into the early stage of product development. As shown, a modular
product architecture of product systems leads to many possible alternatives for
product system configurations, which makes it difficult to identify the ideal com-
bination of modules that leads to financially and environmentally optimized life
cycle properties. Therefore, approaches of mathematical optimization need to be
applied for supporting decision makers during the product development.
The methodology that is developed in this work, in general, is applicable for
various environmental targets that have been discussed previously. However, for
better understanding, it is explained and demonstrated for the reduction of GHG
emissions. For this study, the social aspect of the triple bottom line will not be
included in the calculation models, leaving the focus on the economical and the
environmental aspects (in the sense of eco-efficiency). Considering this exemplary
scope, this work contributes to the reduction of greenhouse gases by providing a
systematic methodology that analyses the most cost-efficient pathways for prod-
uct decarbonization. This methodology helps to develop eco-effective product
systems if the determined environmental target is in line with the goals of the
Paris Agreement.
State of Research and Identification
of the Research Gap 3

3.1 Criteria and Requirements for Optimizing

the Eco-effectiveness of Product Systems

Several prerequisites need to be fulfilled when facing the challenge of supporting

the product development by optimizing the eco-effectiveness of a single product
system or multiple units of this system. These prerequisites describe the frame-
work of the given problem statement and describe the required conditions that a
solution to this problem needs to meet. Some of these conditions focus on the
input data and information that is included in the framework. Other conditions
describe the requirements on the modelling of the problem statement as well as
the algorithm and the solution that is to be found. Furthermore, it is described
which knowledge and framing conditions regarding the product system in ques-
tion have to be available to derive useful results with a suitable approach or
tool.
The prerequisites that are described in this chapter are distinguished between
criteria that a general solution approach has to meet in order to enable correct
solutions on the one hand and additional requirements that help the user to gen-
erate stable and comparable solutions on the other. The criteria described in the
following subchapters are briefly summarized after each paragraph and later used
to evaluate the current state of research. The additional requirements, however,
are not used for the comparison of possible approaches as they do not describe
the solution approach but the product system itself. Therefore, those requirements
do not influence the suitability of an approach regarding the suitability to solve
the given problem statement.

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 63

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_3
64 3 State of Research and Identification of the Research Gap

3.1.1 Criteria to Handle Product System Modularity

It is relevant for a proper solution approach to not just compare different avail-
able product systems and identify the most suitable out of them, but to assist in
creating a new product system that meets the given restrictions to support the
development of future product systems. This requires a shift from the retrospec-
tive assessment of product systems towards an approach that strives to engineer a
product that is designed to meet the targeted limits. In order to be able to propose
new product system configurations, different product modules or components and
their alternatives need to be given as input data for each function of the product.
Those different product modules that are combined for the final assembly of the
product serve as a modular construction system.

 Criteria (C)1: Availability of individual product modules and their alterna-

tives

The availability of many different modules and module alternatives quickly leads
to a high number of possible combinations (see section 2.1). If the combination of
those modules is not regulated in any way, it is possible that, among all resulting
product systems, there are many combinations that are either technically impos-
sible or undesired. These combinations will unnecessarily be part of the solution
space and thus the optimization process. Including such combinations ultimately
makes the search for the best possible solution more complex. In order to reduce
complexity, only module combinations are to be created that represent a valid
product system. This requires the application of a model or an algorithm that
avoids the consideration of certain combinations.

 C2: Avoiding undesired module combinations

It is important that the selected approach covers both the ecological and the
financial impacts of the final product system along its entire life cycle in order
to achieve a higher eco-effectiveness. This requires having both the LCA and the
LCC results available for each product module. These results are used to calculate
the life cycle performance of the final product system configuration by adding up
the results of each module.

 C3: LCA and LCC results for each product module are available
3.1 Criteria and Requirements for Optimizing … 65

However, combining several modules and adding up their individual LCA and
LCC results may lead to wrong results if interdependencies between different
modules occur (see section 3.2.1). If different modules of a product interact and
the specification of one module influences the performance of another module
(e.g. a reduced weight of a module enhances the energy efficiency of the engine),
the static results of LCA and LCC (e.g. for the engine) can be incorrect because
each configuration would have an individual result. This means that those inter-
dependencies between individual modules have to be considered for the overall
results regarding the ecological and financial performance. Hence, a modular
optimization approach of the product system has to meet the requirements of
considering such interdependencies.

 C4: Consideration of interdependencies between combined modules is

possible

3.1.2 Criteria to Handle Use Case Specific Requirements

The optimization of eco-effectiveness can result in the implementation of mod-

ules that require special production processes or infrastructures in its supply chain
or operation. A holistic approach for the optimization of the product system
must consider the investments in or scale-ups of supporting processes that do
not belong to the direct product system. This is required if despite the investment
in new production technologies or infrastructures, the new technologies lead to a
higher eco-effectiveness.

 C5: Consideration of additional production processes or infrastructure for

optimization of product system

As the developed methodology shall also be applicable to not only single prod-
uct systems, but also to multiple product systems (e.g. a fleet of vehicles), it is
required to identify the ideal configuration for not just one product, but also for
multiple units of this product system. Therefore, the applied solution approach
must be flexible in optimizing any amount of product units given the same
framing conditions.

 C6: Optimization of single product system or multiple units of the product

system (product fleet)
66 3 State of Research and Identification of the Research Gap

However, not every product module or measure might be available for every
product system. The optimization algorithm needs to consider these limitations
and must identify the optimal assignation of modules and product units given
these conditions.

 C7: Consideration of possible limitations regarding the availability of

modules

Also, a time perspective needs to be included into the calculation of the results.
Over the time of operation, increasing or decreasing costs for e.g. fuels or energy
for the use stage can influence the choice for the ideal product configuration.

 C8: Consideration of impact from time-dependent price developments

3.1.3 Criteria for Optimization Approach

The described requirements from the problem statement need to be fulfilled by

the optimization approach. Furthermore, the approach needs to be able to manage
and process the given input data and handle the given problem statement in an
efficient way. An additional criterion for the optimization algorithm is being able
to switch easily between the optimization of a financial minimum for a given
environmental limit and a minimum environmental impact for a limited financial
budget. As the target is the optimization of eco-effectiveness, the limit values
need to be absolute figures. Furthermore, the algorithm must be designed in a
way that it can handle extensive product alternatives with many possible module
combinations.

 C9: Enabling both absolute financial and environmental targets as optimiza-

tion objective
 C10: Handling a large number of product alternatives

Next to the interchangeable optimization objective, it is also relevant for the

solution approach to enable an optimization that can focus on different life cycle
stages. For certain applications (e.g. in the automotive industry), it is important
to minimize environmental impacts along the entire life cycle to comply with
absolute sustainability goals. However, at the same time it is necessary to meet
a limitation of a governmental regulation that considers the use stage only. This
means that the optimization algorithm must meet several secondary constraints
3.1 Criteria and Requirements for Optimizing … 67

and needs to be able to separate the individual life cycle stages in terms of
financial and environmental performance.

 C11: Enabling the consideration of secondary constraints regarding different

life cycle stages

A high solution quality, a fast solution time and an appropriate presentation of

the results leads to an informative visualization of the results.

 C12: High solution quality, quick runtime and appropriate visualization of

problem and results

3.1.4 Additional Requirements to Obtain Useful Results

Further requirements and information regarding the product system in question

are required to achieve applicable solutions. These requirements are not used to
evaluate the suitability of a solution approach, but are still helpful to select a
suitable method.
The considered product system that is to be optimized (e.g. a vehicle) needs to
be well known by the product developer in order to be able to apply the optimiza-
tion concept. It is important that the product system and its requested functions
are precisely described to make sure that the algorithm combines the modules in
such a way that all required functions are provided. Ideally, there is a reference
product unit or a predecessor model that can be used as a benchmark for the
new product system. Furthermore, the product architecture and the hierarchical
composition of the product’s components have to be known. The information
regarding the number of modules, their connection and position within the prod-
uct is a crucial aspect to model the product system correctly. In order to increase
the level of detail and the quality of the result, specific details regarding the
components, such as weight, dimensions and material used, is required for the
quantification of the financial and environmental burdens. This information can
e.g. be provided by a bill of materials for the given product or a predecessor
model.
The required information exceeds the narrow boundaries of the physical prod-
uct unit and applies to the production processes and the end of life treatment as
well. For a complete modelling and optimization of the life cycle performance,
the supply chain of the components and their materials also have to be known.
The same applies to the assembly and dismantling processes and the specific use
68 3 State of Research and Identification of the Research Gap

of energy and materials for those steps. For the materials used and the production
processes the corresponding prices need to be available.
In addition to the description of the initial product system, it is relevant
to know which components or modules can be replaced by alternatives. These
alternatives are characterized by e.g. different materials, a different geometry or
different principles of functionality while maintaining the function of the original
product system. These alternative modules and other possible measures during
the production, use or end of life stage form the basis for the alternative prod-
uct configurations of which the optimization model chooses the ideal set. For
the optimization of multiple product units, it is also required to have information
regarding the availability of individual modules and measures. If a certain module
cannot be provided for the entire product fleet (e.g. because of limited material
supply), these boundary levels have to be known to make sure the optimization
algorithm does consider those limitations.
Moreover, clearly defined assumptions need to be determined for the framing
condition of its life cycle, especially with regard to the use stage. A fixed use
stage needs to be defined that describes the time frame of the use stage and a
defined way of utilization. This standardization of the use stage is required to
cover the financial and environmental impacts over the use stage for all product
alternatives in an equal manner. The time frame of the use stage is defined as
a temporal limitation (e.g. 15 years) or as a limited number of operations (e.g.
100,000 km of mileage for a vehicle). The use profile is defined by using compa-
rable and reproducible test cycles that include all relevant actions and represent
the average use profile of the product. For the automotive industry, such cycles
are described in section 2.3 (NEDC or WLTP cycles).
For the evaluation of life cycle costs, the time horizon for the use stage
has to be defined to make suitable predictions for the development of energy
and material prices. Additionally, the financial perspective has to be declared
as the optimization can lead to different results for either the customers’ or the
manufacturer’s point of view (see section 2.2.2).

3.2 Current Approaches and State of Research

As stated in section 3.1, approaches for the optimization of eco-effectiveness

require (1) a sufficient product modularity to incorporate interdependencies
between modules, (2) data input of LCA and LCC for the individual modules
and the final product systems, and (3) a suitable method for the optimization
of the information regarding the eco-effectiveness. To identify relevant literature
3.2 Current Approaches and State of Research 69

that covers all these aspects at the same time, all three areas are searched for
approaches that also overlap with the other fields.
For the field of product modularity, approaches and publications are analyzed
regarding how the introduced modularity is used and if this modularity has a
benefit in terms of optimization of modular products with regard to the LCA
and LCC results. The literature on modular LCAs is analyzed by assessing if the
modular approach is used for optimization of the product system by selecting the
ideal module for each step. The literature focusing on LCA & LCC optimization
is analyzed with regard to modular product systems and product population fleets.
Figure 3.1 shows how the analyzed literature is selected by focusing on
approaches that combine all relevant aspects. On the left, a three staged graph
is shown that shows whether a topic field is considered (dark box) or not (white
box). Every stage adds another prerequisite, starting with product modularity and
modular LCAs to the optimization of LCA and LCC. As only those approaches
are relevant that cover all three aspects, only the lowest route through the graph
is relevant for the literature research. Only for this path all stages are marked as
“yes”, resulting in the overlapping focus area that is also shown on the right hand
side of Figure 3.1.

+ Stage III:
Opmizaon
of LCAs & LCCs

1 1
+ Stage II: Product
Modular 2 modularity
LCAs
5
Stage I:
Product 3
modularity

4 7 6
8
Focus
5 area

6 3 2
4
7
Legend

No Yes 8

Figure 3.1 Focus area for literature review

The current state of the art of these areas of research in the field of life cycle
engineering is described in this subchapter.
70 3 State of Research and Identification of the Research Gap

3.2.1 Product Modularity

Based on the definition of modular products in section 2.1, the methods to divide
a product system into the ideal number of modules and the analysis of the
relations between the individual modules have developed over time.
Umeda et al. (2008) introduce a methodology to determine modular product
designs that fulfill both LCE objectives such as upgradability or recyclability
and a geometrical feasibility of the product architecture (Umeda, Fukushige,
Tonoike & Kondoh 2008). Tseng et al. (2008) evaluate and calculate the liai-
son intensity of components and depict these connections by using liaison graphs.
Then, a grouping genetic algorithm is applied to cluster the components into mod-
ules. The modular product architecture is then evaluated in terms of production
cost and environmental impact. Afterwards, iterative changes to the components
are evaluated to identify “green life cycle engineering”. (Tseng, Chang & Li
2008).
ElMaraghy and AlGeddawy (2013) introduce an analysis of the granularity
of modules in a product system, with a deeper hierarchy of components indicat-
ing the level of the granularity. For a high interchangeability of modules within
product families, the level of granularity needs to be considered. A design struc-
ture matrix is combined with a hierarchical clustering, resulting in a clustering
tree that is used to identify the ideal level of module granularity. (ElMaraghy &
AlGeddawy 2013) The mentioned research mainly focuses on improvements for
manufacturing and assembly of product systems.
Halstenberg et al. (2015) state in their research that a modular product archi-
tecture is also useful for addressing targets of a sustainable product design. Also,
based on a broad literature review, different forms of dependencies between
components have been identified. They form the basis for the clustering of
components into modules. (Halstenberg et al. 2015) Those forms of module
interdependencies are: “Independence and similarity analysis, component position
pattern, assembly dependency, accessibility, cost of reusability, interface openness
and interface design effort” (Halstenberg et al. 2015, p. 604).
Wang et al. (2016) studied the interdependencies between individual compo-
nents of a product system in order to be able to include these interactions into
their approach to enable a green modular design. A prediction model has been
developed that helps to identify how many other components are affected by the
change of one component. A propagation tree is used, providing the tracking of
the probability of influences between models, based on stochastic variables. The
results are then displayed in a dependency matrix between the modules. The indi-
vidual likelihood of a dependency between two components is derived from the
3.2 Current Approaches and State of Research 71

experience of the product designer. The influence of a module on the fulfilment

of the desired product function is also evaluated in order to understand if the
change of a module leads to a reduction of the product’s performance. The inten-
sity of the physical connection between components is analyzed and expressed
by liaison graphs. (Wang, Tang, Yin & Yang 2016).
Mutingi et al. (2017) present a liaison graph for the assembly of a pen. The
nodes of this liaison graph symbolize the different product modules and the arcs
show the liaisons between the modules. The weight of each arc defines the inten-
sity of the liaison (see Figure 3.2). Ideally, the liaisons within the components
of a module are maximized while the liaisons between different modules are
minimized. (Mutingi, Dube & Mbohwa 2017).

1. Button

2. Body
4. Head 5. Tube 6. Ink

3. Cap
Liaison intensity

Figure 3.2 Liaison graph for module connections (Mutingi et al. 2017, p. 473)

The research on product modularity shows that the product components and
the architecture in which the components are structured have been analyzed
in many aspects by various authors. The biggest focus of the studies is on
the optimized grouping of components into modules. Also, the interdependen-
cies between individual components and modules have been considered. The
knowledge of product modularity is often used for the ease of manufacturing
or a simpler mass customization. In terms of sustainable product development,
the results and findings have mostly been linked to life cycle aspects like
upgradability or recyclability.
72 3 State of Research and Identification of the Research Gap

3.2.2 Modular Life Cycle Assessments

Performing LCAs in general is a complicated and cost and time intensive task.
The accumulation of extensive input data and process information, the process
modelling, and the result calculation require time and calculation capacities. Also,
the LCA results need to be interpreted correctly by an expert to derive relevant
knowledge. (Kuo, Smith, Smith & Huang 2016; Otto et al. 2002) Simplifications
to the LCA calculation can be made by the principle of modular LCAs to help
making LCAs more easily applicable to many products and companies.
Figure 3.3 shows how a modular LCA is defined in contrast to a conven-
tional LCA. As defined in section 2.2, a conventional LCA is performed by
modeling the entire product system connecting it to the background data and
aggregating all occurring elementary flows. The classification of the LCI results
leads to the system’s LCIA result. In contrast to this approach, a modular LCA
consists of individual information modules, which are modelled including both
foreground and background processes of those modules. The impact assessment
is done before the entire product system is modelled instead of afterwards, which
leads to reusable LCA modules. After the individual modelling of each mod-
ule, all modules are connected in order to result in the total product system.
(Buxmann, Kistler & Rebitzer 2009).
The final LCIA result of the entire product system is the sum of the individual
LCIA results of each module. The conventional and modular LCA both result in
the same indicator result. The main difference is the order in which the different
processes are modelled and calculated. While the conventional LCA models the
entire system first and then calculates the LCIA result, the modular LCA models
and calculates separated modules and recombines them afterwards. (Buxmann,
Kistler & Rebitzer 2009).
The systematic literature review of Sonego et al. (2018) regarding “the role of
modularity in sustainable design” shows that the general topic of modularization
in life cycle engineering is becoming more and more relevant. Most published
and reviewed articles in this field focus on the aspect of green modularization.
Their goal is to provide solutions for the challenge of clustering components into
modules in a way that environmental burdens are minimized or life cycle concerns
such as reuse or recycling are improved. Another focus lies on the implementation
of modularization following the principle of “Design for X”. Only four out of the
145 identified publications connect the product modularization to the topic of life
cycle assessment. (Sonego, Echeveste & Debarba 2018).
Jungbluth (2000) introduces modular LCAs by dividing the initial product sys-
tem into clearly separated modules and assessing the life cycles of these modules
3.2 Current Approaches and State of Research 73

Figure 3.3 Comparison of conventional LCA and modular LCA (Buxmann et al. 2009,
p. 96)

individually. In his work, Jungbluth (2000) uses the example of meat produc-
tion and separates the life cycle into five independent modules: (1) agriculture,
(2) processing and distribution, (3) transport, (4) packaging and (5) consump-
tion. For these modules, individual LCAs are calculated. This approach results
in individual LCAs for each module of the product system. The final LCA of
the entire product is then calculated by recombining the modules and adding up
their LCA results. The advantage of this method is the calculation of LCAs for
modular product systems where several modules can be replaced by alternative
modules. As every module has its own LCA result, many different product sys-
tems are configured by adding up the module results without having to perform
new LCAs for every possible product system. To assess alternatives regarding
e.g. transport, packaging, or conservation, only the corresponding module needs
to be adapted and afterwards combined with the other modules again. (Jungbluth
2000, p. 249 f.) Modular LCAs help to evaluate the impacts of different product
74 3 State of Research and Identification of the Research Gap

parameters and show the differences between different product system alterna-
tives. These features make the approach extremely helpful when a wide range
of possible product configurations is assessed. (Jungbluth, Tietje & Scholz 2000)
In comparison, a conventional LCA is not flexible enough to calculate the same
large amount of different LCAs for differently configured product systems in a
comparable time span, as every variant needs to be modelled individually.
According to Jungbluth (2000), modular LCAs lead to new challenges, as
the separation of a product system into modules can be accompanied by prob-
lems regarding allocations. This occurs when an environmental impact is linked
to more than one product module. This environmental impact then has to be
attributed to one module and separated from the other. As a consequence, changes
in the isolated module cannot be considered anymore regarding that specific
environmental impact. Furthermore, modular LCAs lead to a higher effort for
the interpretation of the results as often not all theoretically possible module
combinations result in technically feasible products. (Jungbluth 2000, p. 249 f.)
Otto et al. (2002) have shown that the complexity of integrating LCAs into the
product development phase, especially for product families, can be reduced by
identifying common modules within all products of the product family. Assessing
these modules once but considering their results for all products reduces the
LCA effort. (Otto et al. 2002) Rebitzer (2005) has proven that the results of
a modular LCA for a product system are equivalent to the results of a regular
LCA. While the method of calculation differs from the conventional approach, a
modular LCA is in line with the ISO 14040 and 14,044 standards. Rebitzer (2005)
states that modular LCAs are a suitable tool to analyze product alternatives or
product variations with less complexity.
Dose (2005) shows in her research that modular LCAs can be affected by
interdependencies between modules as changes between module alternatives can
affect the whole product system. Dose (2005) introduces exchangeable life cycle
inventories (LCI) of modules in order to be able to still use modular LCAs for
product systems that are sensitive to module interactions. These interchangeable
LCIs of the modules are clearly separated by cut-offs from their surrounding
modules in the product system, leading to parametrized modules with interfaces
for a modular recombination. Those separated LCAs of modules help to calculate
LCAs for different product systems much faster. (Dose 2005).
The combination of LCAs with a modular product design has also been studied
by Reccioni et al. (2007) in 2007. According to their research, “the modularity
concept implies many aspects which allow some simplifications on the application
of the Life Cycle Assessment (LCA) methodology” (Reccioni et al. 2007, p. 54).
For product systems with a high modularity, the change of a single module does
3.2 Current Approaches and State of Research 75

not affect the rest of the product system. Hence, the changes to the LCA result
only refer within this module. Thus, if a module is exchanged, only the new
module has to be assessed. Reccioni et al. (2007) state that LCAs of independent
modules decrease the required calculation time and the required amount of data
as less information needs to be processed.
In a review for sustainable product design, Chio and Chu (2012) have identi-
fied that “most sustainable design methods developed in the past failed to address
the interdependencies among different stages in a product’s life cycle. This defi-
ciency may result in biased estimation and wrong decisions.” (Chiu & Chu 2012,
p. 1268). This leads to a need for approaches that integrate “the scopes of product,
process, system, and ecosystem, while balancing conflicting product development
perspectives.” (Chiu & Chu 2012, p. 1268). Such an approach can e.g. be imple-
mented by applying modular LCAs into the product development phase (Sonego
et al. 2018).
For the development of a predictive eco-design process for the product devel-
opment phase, Kuo et al. (2016) use product modules as attributes of the product
system’s LCA. This helps to reduce time and cost efforts as well as the required
amount of calculated LCAs. The predictive LCA approach incorporates modular
LCAs as well as graph modelling and graph searching approaches like depth first
search. Based on a similarity threshold, the individual components and modules
are grouped. Their LCA is predicted based on the LCA results of previous similar
designs. (Kuo et al. 2016).
Kim and Moon (2019) introduced an approach to utilize modular product
architectures to optimize the product design with regard to the product recovery
at the end of life stage of a product’s life cycle. The target is to enable a modular
product design in such a way that each module does not have to be dismantled
further to the level of sub-components or material to be recycled. Thus, the mod-
ule is the smallest required unit for dismantling and further steps can be avoided
to reduce costs and environmental impacts. (Kim & Moon 2019).

3.2.3 Optimization of LCA and LCC

Azapagic and Clift (1998) first introduced the combination of LCA and LCC
analyses of a product system and the joined optimization of both dimensions
for an optimized product system. Their research is based on a multi-objective
linear programming approach to optimize multiple environmental or financial
objectives of the product’s life cycle. (Azapagic & Clift 1998; Azapagic & Clift
1999) Since then, the research field for the optimization of both environmental
76 3 State of Research and Identification of the Research Gap

(LCA) and financial (LCC) aspects of a product’s life cycle has been analyzed and
studied by many authors, covering different subareas. The results provide multiple
approaches to identify product systems with an optimized ratio of lifetime costs
to environmental impacts.
The literature review of Miah et al. (2017) categorizes the different approaches
of LCA and LCC integration into six separate types: independent LCA and LCC
(Type I), independent LCA and LCC as part of an overarching framework (Type
II), independent LCA and LCC analysis integrated by MCDA (Type III), opti-
mization of LCA and LCC analysis (Type IV), environmental LCC (Type V),
and eco-efficiency methods (Type VI) (Miah, Koh & Stone 2017, p. 850). For
the research of this work, the types IV and VI are most suitable in terms of
methodology and desired results.
In approaches that fall into category IV, the optimization of LCA and LCC is
typically designed by an objective function that is to be minimized or maximized
with respect to boundary conditions and constraining functions. The solution
space becomes extremely large due to a high number of possible outcomes.
Therefore, optimization algorithms like (non-)linear programming are applied to
the problem statement. Across the 32 identified approaches of this type, Miah
et al. (2017) found a broad spectrum of applications, with most studies being
performed for the analysis of buildings and energy systems. In these approaches,
various algorithm types like multi-objective linear programming, mixed integer
linear programming, genetic algorithms, or multidisciplinary design optimization
were utilized. The identified drawbacks for these approaches are the required
effort for system modelling and the required information for the implementation
of the algorithm into a mathematical software. The 19 studies that form type VI
of the LCA and LCC integration show a similar distribution of various fields
of application. The central difference between the methods of type VI is their
definition and calculation of the eco-efficiency index. (Miah et al. 2017).

LCA and LCC optimization for product systems

As defined in section 2.4, the optimization of one or more parameters of an objective
function aims to identify the ideal set of parameters to either maximize or minimize
the desired value. Such optimization approaches can also be applied to the use cases
of LCA or LCC.
An exemplary adaption of the LCA and LCC-oriented optimization for the field of
buildings is described by Islam et al. (2015). In this approach, possible combinations
of modules such as alternative walls, floors and roofs regarding different designs
and materials are assessed and optimized. A single-objective optimization algorithm
minimizes either the life cycle costs or the environmental impacts. Within this study,
3.2 Current Approaches and State of Research 77

a total of 31 alternative building configurations are evaluated while interactions

or interdependencies between modules are neglected. (Islam, Jollands, Setunge &
Bhuiyan 2015).
The study of Herrmann et al. (2014) regarding a “structural” LCA approach
describes alternative life cycle pathways of a product system as variations or alter-
natives of individual processes. A “structural LCA” is defined by Herrmann et al.
(2014) as the systematical addressing of many different variations or alternatives.
These alternatives lead to an extremely high number of possible pathways due to the
high number of possible alternative combinations. The introduced example is a set
of 1.09 ∗ 1012 different product system configurations (or alternative life cycle path-
ways). However, not all combinations are technically feasible. The regular approach
of LCAs, where only a few variants are analyzed, fails to analyze these pathways in
a short time as every time a completely new LCA has to be performed. The structural
LCA approach on the other hand manages to increase the possible amount of LCAs
and provides new possibilities for analyzation and optimization. This optimization
is either focused on single-objective or multi-objective functions. The individual
pathways are organized by a unique ID number within a structural table. The com-
plexity increases if interdependencies of different pathways occur. Herrmann et al.
(2014) recommend statistical software tools to cover these interactions, but did not
include them in their model. (Herrmann et al. 2014).
Buchert et al. (2015) describe an approach in which decision trees are used
to model the hierarchical structure of the product system. Such a decision tree
visualizes the alternatives that occur at each new step of the product system’s life
cycle. Each branch of the decision tree is quantified by the environmental impact of
the given alternative. The decision tree resembles the “value creation network” of the
product. The ideal path for a sustainable product (covering LCA and LCC) through
this network can be identified by applying a multi-objective decision algorithm to
it. Buchert et al. (2015) show that the effort of modelling and the solution space
grow quickly with every new hierarchy level in the decision tree as the number of
possible combinations grows. A higher level of detail for each step also increases
data demand and effort for modelling and calculation of the ideal solution (Buchert,
Neugebauer, Schenker, Lindow & Stark 2015).
Nadoveza et al. (2013) present a link from conventional LCAs to LCAs based
on graph theory to reduce the required time for decision making. The choice of
the graph theory for modelling the algorithm is based on the advantages regarding
visualization of the problem statement and solutions as well as on the broad avail-
ability of different optimization algorithms for graphs and networks. The approach
uses those advantages by modelling a product system and their life cycle factors as
a graph connecting the interfaces of each factor with edges. The relations between
78 3 State of Research and Identification of the Research Gap

the nodes are represented by interdependencies that occur between the individual
life cycle factors. A change of one factor and the resulting effect on another fac-
tor can be visualized within the graph. (Nadoveza, Koukias, Karakoyun & Kiritsis
2013) The approach of Nadoveza et al. (2013) “provides the capacity to set up the
goal and scope of analysis (e.g. carbon footprint below a threshold), and the system
may then find all the possibilities and opportunities in order to achieve that goal”
(Nadoveza et al. 2013, p. 413). The outlook of their study indicates the necessity to
expand the graph based LCA optimization by a financial dimension like life cycle
costing. (Nadoveza et al. 2013).
The idea of analyzing and optimizing the LCA result of a product system with the
tools of graph theory is further described by Jahandideh et al. (2015) “to overcome
the complexity of traditional Matrix-based analysis” (Jahandideh, Aminikhang-
hahi, Salehnia & Muthukumarappan 2015, p. 1). For lignin derived chemicals, a
network for the current and possibly alternative production processes is designed.
The goal is to optimize the environmental and economic dimensions of the product.
The connecting edges of the graph are weighted with the conversion yield of the
corresponding production process as well as the environmental impact of that path.
If a certain combination of chemicals is not feasible, the network stops at the last
possible node and illogical combinations are eliminated. As the graph is designed
following the physical material flows, the model is an acyclic graph with a single
source and a single destination node offering multiple pathways in between. (Jahan-
dideh et al. 2015) The decision problem is formulated as a minimization problem
to optimize the production process with respect to a design path with reduced envi-
ronmental and economic burdens. This minimization problem is programmed as a
shortest path problem through the network and solved with a modified version of
the Dijkstra algorithm. Maximizing the profit on the other hand leads to a problem
formulation equivalent to the identification of the longest path through the net-
work which resembles a NP-complete problem regarding the complexity class. A
multi-objective analysis is performed leading to trade-offs between both dimensions
to find a solution with minimal environmental burden and a high financial profit.
(Jahandideh et al. 2015).
A further approach to enable the evaluation of an extensive number of product
configuration alternatives for modular products is presented by Steubing et al. (2016)
who focus on streamlining the LCA calculation in order to reduce the calculation
effort. The analyzed example describes alternative process paths to generate heat
from biomass. The process chain contains five consecutive steps with two to four
alternatives per step, resulting in 14 process modules that can be connected to 144
different process alternatives (see Figure 3.4). The idea of the modular LCA is then
to only calculate the LCAs for the 14 modules and recombine them to the 144
3.2 Current Approaches and State of Research 79

options instead of directly modelling all alternatives. A drawback of this approach

is the computational structure of LCAs as it relies on process-process links with
every input of a process coming from a certain upstream process. Therefore, regular
LCI databases are not suitable for large scenario evaluations. Hence, in order to be
able to calculate all 144 options in a modular way, copies of upstream processes for
each module are required to cover each individual path. For the 144 options of the
example, this leads to 212 required processes (see Figure 3.4). (Steubing, Mutel,
Suter & Hellweg 2016).
Figure 3.4 a) shows the 14 different modules in five different hierarchical steps.
The black arrows represent a selected path while the dotted arrows show all other
possible connections between the modules resulting in 144 alternatives. Figure 3.4
b) depicts the further growth of process modules if specific upstream paths are con-
sidered. LCI modules of intermediate process steps need to be cleared of their own
supply chain impacts to avoid a double counting of processes, as this information
is provided by the previous module. The full enumeration of these paths results in
212 required modules. (Steubing et al. 2016).
The process system in Figure 3.4 is used by Steubing et al. (2016) as a model
for an optimization problem with the target to minimize the environmental burdens.
The organization of the input data is done using a “module-product matrix” that
contains the environmental information for the possible alternatives of the product
system and links input amounts with outputs for each process. In order to link
every possible value chain with unique module configurations, a square matrix is
needed. Steubing et al. (2016) use a recursive depth-first graph traversal algorithm
to identify all possible value chains that are offered by the module-product matrix,
as they cannot be directly identified from the matrix. The final optimization model
is then based on a generic linear programming algorithm that optimizes the product
system based on the data of the matrices. (Steubing et al. 2016).
Ameli et al. (2017) propose a methodology for the optimization of a product
system with trade-offs between life cycle costs and environmental impacts. Their
approach is optimized for typical product design problems of products with many
components and many alternatives for each component (e.g. automotive or airplane
manufacturing). As an example, a product containing 50 different components is
described (while a typical automobile has up to 10,000 parts) where every component
has three alternatives. This leads to a total of 717 ∗ 1021 options with a required
LCA and LCC evaluation for every single option. (Ameli et al. 2017).
Figure 3.5 shows the data structure for the alternative product designs by Ameli
et al. (2017) that is used for the optimization. The optimization problem of the
approach by Ameli et al. (2017) is designed in such a way that the environmental
80 3 State of Research and Identification of the Research Gap

a b
Module Module Module Module
A1 A2 A1 A2

Module Module Module B1- B2- B3- B1- B2- B3-

B1 B2 B3 A1 A1 A1 A2 A2 A2

Module Module C1- C2- C1- C2- C1-

C1 C2 B1-A1 B1-A1 B2-A1 B2-A1 …

Module Module Module Module

…
D1 D2 D3 D4

Module Module Module

…
E1 E2 E3

Figure 3.4 Representation of alternative module combination options (based on Steubing

et al. 2016, p. 512)

burden shall not surpass a defined limit value while the financial effort is to be min-
imized. While the environmental impacts are assessed over the entire life cycle, the
costs cover only the expenses of certain life cycle stages which are directly related
to a single component (e.g. material prices, transport costs, end of life costs). Costs
of the use stage are not considers, as well as the influences of the components on
the use stage costs are not considered. The optimization model is based on linear
integer programming, where the objective function is the cost minimization and the
constraining functions define the environmental threshold value. The result shows
which alternative of each component should be selected to form the product system.
By changing the limit value for the environmental impact, the financial effort of the
product system either increases or decreases. The most cost-efficient configuration
is selected without setting an environmental limit at all. By changing the objec-
tive function to the environmental impacts, the most cost-efficient configuration
complying with the given emission level is identified. (Ameli et al. 2017).

Automotive focused approaches of LCA and LCC optimization

Research regarding LCA and/or LCC results and their optimization for multiple
products, such as a fleet of automobiles, has been done by various authors like
Reichmuth, Lutz, Manley & Keller (2013), Garcia, Gregory & Freire (2015), Ercan,
Zhao, Tatari & Pazour (2015), Onat, Kucuvar, Tatri & Zheng (2015) and Lemme,
Arruda & Bahiense (2019). Their common focus is the analysis of the environmental
and financial impacts of automobiles and the optimization of the fleet composition
in terms of sustainability goals.
3.2 Current Approaches and State of Research 81

Figure 3.5 Design alternatives for product design (based on Ameli et al. 2017, p. 2451)

Reichmuth et al. (2013) and Garcia et al. (2015) both use a dynamic fleet model
that, based on an initial fleet composition, considers annual vehicle sales and the
amount of replaced vehicles for the fleet transition. The research objective of these
authors is to come up with the best combination of up to six different types of
powertrains for a fleet in the United States for the reduction of greenhouse gases
in the use stage (Reichmuth et al.(2013)) or greenhouse gases over life cycle in
Portugal (Garcia et al. (2015)). Both studies neither include a cost perspective nor
82 3 State of Research and Identification of the Research Gap

an optimization model. The result calculation is based on simple comparisons of

scenario. (Garcia et al. 2015; Reichmuth et al. 2013).
The authors Ercan et al. (2015), Onat et al. (2015) and Lemme et al. (2019) present
methodologies that use mathematical optimization tools to identify the ideal fleet
composition in order to achieve environmental targets. Ercan et al. (2015) developed
an approach to minimize both life cycle costs and environmental impacts based on
a multi-objective linear programming algorithm. The methodology is shown as an
example for a bus fleet in public transport. The results can be adapted to different
budget limitations or emission reduction targets. For a fleet of 100 busses, six dif-
ferent powertrain options are considered. For different driving cycles and financial
or environmental constraints, the ideal fleet composition with regard to powertrain
shares is calculated. (Ercan et al. 2015) Onat et al. (2015) follow a similar approach,
considering seven powertrain options for a multi-objective optimization in terms of
financial, social, and environmental dimensions. Their case study focuses on pas-
senger cars in the USA and offers different scenarios regarding criteria weighting
and influencing parameters such as electricity supply. The calculated results show
how intense the individual powertrains should be used to reach the given sustain-
ability targets. (Onat et al. 2015) The approaches by Ercan et. al (2015) and Onat
et al. (2015) do not consider configurable product systems based on a modular
product architecture. Thus, the possibilities to create new product system variants
that may lead to lower emission levels are limited. Further, no interdependencies
between different component selections are considered, making it impossible to
detect unwanted effects for certain configuration options.
Lemme et al. (2019) apply their methodology to an example of small-scale
car sharing, where three different types of powertrains and their infrastructure are
considered. The analyzed dimensions cover financial and environmental aspects
which are optimized with a single- and a multi-objective linear programming model.
The results show the trade-off of financial and environmental burdens between
conventional and electric powertrains and the significance of the infrastructure.
(Lemme et al. 2019) The presented approach lacks the same flexibility for new
product variants due to missing product modularity like the research of Onat et al.
(2015) and Ercan et al. (2015). Furthermore, the optimization approach of Lemme
et al. (2019) is not explicitly designed to handle a large number of combinations,
but is designed to a smaller example.
The presented research regarding the analyzation and optimization of sustain-
ability aspects of multiple products such as product fleets differ greatly from the
research regarding the optimization of the modular product architecture of a single
product. The approaches for product fleets do not consider modular product systems
and the interdependencies between the modules. Typically, the fleet optimization
3.3 Identification of the Research Gap 83

approaches do not focus on product development, but on the composition of avail-

able products. The considered number of variations or alternatives with only three
to seven different powertrain options is much lower than the possible options that
are considered for a modular product system covering the entire life cycle.

3.3 Identification of the Research Gap

The research field regarding the optimization of the product configuration during
the product development phase with focus on the environmental and financial
characteristics has been studied by many different authors as presented in the
previous subchapters. The high diversity of published approaches, methodologies
and frameworks covers a broad spectrum of different purposes within this large
area of research. Many studies explicitly focused on detailed problem statements
or a limited field of application.
The current state of research regarding product modularity shows that the mod-
ularity of a product system and its hierarchical architecture is a well-established
concept. Many authors have built their concepts on the knowledge of interdepen-
dencies between modules which occur when a change to a module affects other
modules. These interdependencies have been categorized into different types of
interaction (e.g. distance modularity or bridge modularity) and visualized (e.g. by
using liaison graphs). Due to this information, the most centralized modules as
well as the modules that show no or little interaction with the rest of the prod-
uct system are identified. Based on the concept of modular interactions, many
approaches have been developed to consider these interdependencies in the prod-
uct architecture to minimize the mutual influences by designing products with
clearly separated modules for each function. The main driver behind this research
is the optimization of product families, remanufacturing or easier upgradability.
The concept of product modularity has been transferred to LCAs for product
systems by previous authors. Such modular LCAs help to simplify and speed
up the calculation of LCAs for extensive product variants while still leading to
results that are in line with the ISO norm 14,040. The individual assessment
of modules and module alternatives and the following recombination leads to
fewer required LCAs than possible module combinations exist. This, in return,
allows product developers to compare large numbers of design alternatives with
a reduced effort in data analysis and calculation time. The challenges of mod-
ular LCAs are described as the correct separation of the product system into
modules, the right allocation of material or energy flows to the modules and the
84 3 State of Research and Identification of the Research Gap

handling of technically illogical module combinations. The principle of modular

LCAs is especially suitable for products with a high modularity and no or little
interdependencies between modules. However, when interdependencies between
modules become more relevant, a simple application of modular LCAs leads to
wrong results. Currently only very few methods are published that deal with
module interdependencies in modular LCAs.
The further development of modular LCAs and their LCCs has been
described by various authors, often going into different directions regarding their
approaches and targets. The optimization of product fleets focuses on dynamic
fleet models with the target to identify the ideal product composition, but without
considering extremely large numbers of variants, interdependencies, or improve-
ments for product development. For single product systems, often multi-objective
algorithms are used to optimize the product configuration. Optimization using the
graph theory has been introduced by only a few authors. Use cases that provide
large numbers of possible configuration alternatives are either structured with
tables using unique ID numbers for the variants, module-product matrices, or
decision trees. The consideration of interdependencies only occurs occasionally.
The overview of the current research shows that the general principles of
product modularity have been included into the calculation of LCAs while the
interdependencies between modules and their influence on the result are yet
rarely incorporated into LCA modelling. Thus, these aspects do not find sufficient
recognition during the optimization of the eco-efficiency or eco-effectiveness for
products and product fleets.

Evaluation of identified publications

The analyzed approaches in the identified publications need to be assessed in terms
of the formulated criteria and challenges to the problem statement of this work (see
section 3.1) in order to determine the suitability of each approach to the introduced
criteria. This assessment is used to evaluate which aspects have already been dis-
cussed and which have not yet been covered in the available methodologies. Table
3.1 summarizes the criteria and challenges for the evaluation that are introduced in
section 3.1. The comparison of the criteria with the identified approaches is evalu-
ated using a scale of three possible outcomes: either the given approach fulfills the
considered criteria completely (●), covers the criteria, but does not fulfill all require-
ments ( ( )) or does not consider these criteria at all (◯). The overall suitability of
an approach to all criteria of this study’s problem statement is then indicated in five
categories, ranging from a total suitability (●) in 25% steps down to no suitability
at all (◯).
3.3 Identification of the Research Gap 85

Table 3.2 gives an overview of the evaluation of the publications for the opti-
mization of LCA and LCC of product systems regarding the criteria of Table 3.1.
The selected studies that are benchmarked against the criteria are separated into
two different types comprising either selected representative studies that focus on
the optimization of a single product system or on the optimization of a product
fleet. As some approaches by different authors resemble each other, one author is
selected in table 3.2 to represent all approaches that are based on the same prin-
ciples. The twelve criteria are also grouped as they form three different subtypes
of required aspects. These subtypes are the criteria regarding the product system
architecture and its modularity, the criteria regarding the specific aspects of the use
case and problem statement and the criteria concerning the type and functionality
of the applied optimization algorithm.

Table 3.1 Criteria for optimization of eco-effectiveness for product systems

Requirements from product system architecture
C1 Availability of individual product modules and module alternatives
C2 Avoiding undesired module combinations
C3 LCA and LCC results for each product module
C4 Consideration of interdependencies between combined modules
Requirements from problem statement and use case
C5 Consideration of production processes or infrastructure for optimization of
product system(s)
C6 Optimization of a single product system or multiple units of the product
system
C7 Consideration of possible limitations regarding the availability of modules
C8 Consideration of the impact from time-dependent price developments
Requirements for optimization approach
C9 Enabling absolute financial or environmental targets as optimization objective
C10 Handling a large number of product alternatives
C11 Enabling the consideration of secondary constraints regarding different life
cycle stages
C12 High solution quality, quick runtime and appropriate visualization of
problem and results

For the criteria “product system architecture”, all studies that focus on single
product optimization cover most relevant aspects. Especially the consideration of
modular products is implemented in every study. Also, the criteria with regard
86 3 State of Research and Identification of the Research Gap

to avoiding unwanted module combinations and both LCA and LCC calculation is
broadly represented. Still, not all studies apply their methods to LCC (e.g. Nadoveza
et al. (2013), who only focus on LCA optimization). Only the aspect of interdepen-
dencies between modules and their influence on the LCA and LCC results has
hardly or only partly been covered. The exception is the publication by Steubing
et al. (2016), who also cover this aspect. The three selected studies that repre-
sent the research for the optimization of product fleets, however, are very different
regarding the criteria for product system modularity. As these studies focus on a
different application, product modularity, recombination of modules and occurring
interactions are not covered by these methodologies at all.
The difference between the studies for single or multiple-product optimization
continues for the criteria of required application for multiple products. The studies
that focus on just those aspects fully or at least partly cover all required aspects. The
adaption of their optimization models is explicitly designed to integrate multiple
products into their calculation. Furthermore, they offer the theoretical opportunity
to expand these models to also include aspects such as infrastructure or production
processes as well as module limitations. In comparison to these approaches, the
single product focused methods are less focused on these criteria.
A more diverse result is found in the analysis of the applied optimization
algorithms. For all studies that consider financial and environmental aspects in
general, switching between the objective values is possible. As most methodologies
rely on multi-objective optimization, different weightings between the dimensions
can easily be selected. Theoretically, those algorithms can also incorporate abso-
lute threshold values for the financial or environmental aspects that must not be
surpassed, enabling an optimization of not only eco-efficiency but also of eco-
effectiveness. The handling of many possible module combinations (for a single
product unit) and the combination of many product units into a fleet (multiple prod-
ucts), is generally possible with the chosen types of algorithms as they can search
and optimize large numbers of options. The limitation of the number of alternatives
is more often set by the type of data storage and data provision to the algorithm.
The data management of decision trees or graph theory networks on the contrary is
more flexible than the module-product matrices.
Additional constraints, e.g. simultaneously staying below limit values in multiple
life cycle stages, have not been considered in the present studies. Theoretically,
those constraints can be integrated into the approaches that allow assessments not
only for the whole product life cycle, but also for the individual life cycle stages.
The easiest way to include such constraints is done by approaches that use graph
theory or methods with the focus on multiple products because they also consider
use stage emissions from e.g. vehicles. The suitability of the applied optimization
3.3 Identification of the Research Gap 87

Table 3.2 Evaluation of published approaches for LCA and LCC optimization

algorithm is evaluated considering the provided results and their visualization. As

most algorithms are based on linear programming, the results are plain numbers.
The methods based on graph theory, however, express the results as the shortest
paths through networks, enabling a visualization of the results more easily.
88 3 State of Research and Identification of the Research Gap

Identified research gap

The analysis in Table 3.2 shows that the selected authors have contributed impor-
tant aspects to the general goal of the optimization of a product’s eco-efficiency
and eco-effectiveness. The most promising approaches have been introduced by
Jahandideh et al. (2015) and Steubing et al. (2016) as both approaches consider the
most important and complex aspects regarding the introduced criteria. Still, these
two approaches lack either the integration of interdependencies between modules
(Jahandideh et al. (2015)) or the consideration of financial aspects (Steubing et al.
(2016)). Additionally, both approaches do not cover multiple products and therefore
cannot be applied to the ideal composition of product fleets.
The analysis of the total fitness of the approaches of each study shown in table 3.2
indicates that, despite the high diversity of methods, there is no approach that incor-
porates and unites all features that are required to satisfy all the requirements of
the introduced criteria. While every author shows a new and suitable approach for
selected aspects, none of these approaches fully meets all the required criteria that
are fundamental to provide a holistic solution to fulfill all described criteria. To
fulfill all criteria at the same time, a new approach is required that covers all fields
of interest. Compared to the existing approaches, the following aspects need to be
included or strongly improved:

• Sub-targets for life cycle stages:

When optimizing a product like a vehicle, the emissions of single life cycle
stages like e.g. the use stage are relevant, as the use stage itself is also regulated by
legislation. Thus, an ideal solution cannot only focus on total life cycle emissions
as such a solution might end up with emissions during the use stage that exceed
the allowed limit value. Therefore, it must be possible to not only minimize the
total life cycle emissions, but also to stay within additional limits for single life
cycle stages at the same time.
• Module interdependencies:
The approach must consider interdependencies between modules to make sure
that the effects on the whole product system are included that may occur when
changing a single module.
• Visualization of results:
A proper visualization of the results is helpful in order to be able to integrate
the findings into the product development. The visualizations are helpful to
understand and explain the identified solution and make the derived knowledge
accessible for the user of the approach and further decision makers along the
process of product development.
• Consideration of time-sensitive influences:
3.3 Identification of the Research Gap 89

A dynamic consideration of time-sensitive influences on the life cycle costs

must be added to the existing approaches. The variation of energy costs along
a use stage is one example of such time-sensitive influences. Considering time-
sensitive influences assures finding solutions that may not be the most cost-
efficient solution today, but might be in future when the costs of e.g. fossil
fuel-based energy sources will strongly increase.
• Flexible optimization model:
The optimization model must be adaptive with regard to the amount of product
units that are included in the optimization process, making it possible to choose
if only a single product system is considered or multiple products. Furthermore,
when selecting multiple products, it has to be possible to have a flexible number
of products entering the model. The user must have the possibility to choose
how many product systems they want to include, ranging from a minimum of
two units to up to multiple million product units.
• Data management concept:
For all those described new aspects, a suitable data management concept is
required that organizes all data input and provides all relevant information to
the optimization model. Such a data management must be flexible regarding
changing input parameters, such as varying energy prices or varying CO2 eq. per
kWh electricity. The data management system must also have an interface with
the optimization algorithm to transfer the input data into the simulation.

In conclusion, a new approach is required that combines the strengths of the methods
for single and multiple- product optimization. This approach needs to be extended by
a consideration of module interdependencies, module limitations, time-dependent
cost influences, life cycle stage constraints and a comprehensive result visualization
of the optimized eco-effectiveness of the product system(s).
Concept for the Optimization
of Eco-effectiveness of Product 4
Systems

The optimization of eco-effectiveness of a single unit or multiple product system

units requires a holistic concept. Such a concept has to be in line with the criteria
introduced in Section 3.1 and it also has to cover the process from generating
input data, data management and data optimization to the visualization of the
results. This concept will be introduced step by step in the following subchapters.
First, the requirements based on the identified research gap are defined. Next, the
general framework of the methodology is described and followed by the selection
of the optimization approach. Afterwards, the concept is explained in terms of
network modelling, input data provision, data optimization using graph theory
and interpretation of results. As already stated in chapter one, the methodology is
closely developed to particularly meet the specific requirements of the automotive
industry.

4.1 Concept Requirements

In Section 3.1, criteria and requirements have been introduced that are required
for a holistic solution approach for the optimization of a product system’s eco-
effectiveness. Section 3.3 describes the current research gap, where multiple
aspects have been identified that have to be improved to result in an approach that
allows a comprehensive optimization of product systems regarding eco-effective
targets. Those criteria need to be transferred to specific requirements (R):

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 91

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_4
92 4 Concept for the Optimization of Eco-effectiveness of Product Systems

• R1: The entire optimization approach consisting of product system modelling,

data management and mathematical optimization has to be constructed around
the concept of modular product architectures. These modular products have
to be separable and reconfigurable with alternatives in all technically possible
ways to ensure all valid options to be considered.
• R2: The approach does not only have to model the product system as a mod-
ular architecture, but it also has to be able to move through this modular
structure in an intelligent way. That means that although multiple options per
desired function are available, only one is selected. And it is also required
that the algorithm does not only select a module, but remembers which mod-
ule exactly has been chosen for each desired function for the entire product
system structure. This is fundamental as interdependencies can occur when
changing modules.
• R3: The modular product structure has to be modelled in a flexible way. It has
to be possible to add or replace single module alternatives or entire module
families without making the entire model invalid.
• R4: Every available choice from all module alternatives for the optimization
algorithm has to be characterized individually with attributes that are only
linked to that specific module. These attributes include the environmental
impact (GWP) calculated by an LCA, the costs (e) of the TCO calculated by
LCC and the availability (%) of this module for being included into a product
fleet. This information has to be fixed to the module they describe, but it also
has to be flexible in order to be changed if influences from interdependencies
occur or if time-dependent influences such as varying costs arise.
• R5: The approach has to be completely flexible regarding the amount of prod-
uct systems that are to be optimized. The variety of product systems ranges
from a single product unit to a fleet of product systems with millions of indi-
vidual units. When optimizing product system fleets, it is required to not only
provide average solutions for the entire fleet, but also individual results for
every single product unit within the fleet.
• R6: The optimization approach needs to be flexible regarding the targeted
value that shall be optimized. It has to be possible to switch between an
optimized (reduced) GWP while meeting a financial limit or reducing the costs
while meeting an environmental limit. Both dimensions have to be available
for optimization to provide ideal decision support for many applications in
product development.
• R7: The optimization approach has to be set up in a modular way, so that
not only results for entire product systems are calculated, but also for smaller
segments of the product system. This is required to e.g. represent different life
4.2 Framework for the Optimization Concept 93

cycle stages, such as only the production or the use stage. It has to be possible
to reach a target value over the entire product system while meeting sub-targets
for selected product system sections. In this way, e.g. a fleet emission value
over the entire life cycle is calculated while meeting emission regulations
during the use stage for vehicles.
• R8: The results that are created by the algorithm have to be interpretable and
visualized to assist the user in the decision-making process. The results have
to be available as fleet averages, but also for every single product unit indi-
vidually to be processed for visualization. The choices made by the algorithm
have to be presentable in a comprehensible way by showing each selection
for every module in the structure network of the product system. It has to be
understandable how each product unit within a fleet is constructed and how
the fleet in total performs. Considering individual data sets for every product
unit, the results also need to be plottable in a scatter plot to visualize the
ratio of costs to GWP for every product unit and also the fleet average. When
visualizing all possible combinations of module alternatives, the efficient fron-
tier can be used to quickly identify the ideal product configuration for single
product optimization.

4.2 Framework for the Optimization Concept

A framework is required that describes the linkages between the individual steps
of product system modelling, data gathering and optimization to incorporate all
these requirements into a coherent approach. In the following sections, the general
framework is introduced and the selected optimization approach described.

4.2.1 General Framework

The high number of possible product system configurations arising from modular
products requires modular input data of LCA and LCC and a modular optimiza-
tion. For such product systems, optimization has to be possible for single or
multiple product units and consider the possible interdependencies between com-
ponents and modules. This optimization needs an overarching framework that
connects the individual aspects to a workflow that covers all the aspects and crite-
ria that are introduced in Section 4.1. The framework for the developed approach
in this work is shown in figure 4.1. That framework also defines the structure of
94 4 Concept for the Optimization of Eco-effectiveness of Product Systems

this chapter, as the following subchapters are based on the introduced steps of
the framework.
The framework describes five required and consecutive steps that altogether
form the optimization concept. The five steps include (1) the definition of the
product system and its possible variations due to alternative modules and mea-
sures, (2) the generation of the combination specific LCA and LCC data, (3) the
arrangement and management of this specific data into the modular product sys-
tem structure. Finally, these data are transferred into the optimization algorithm
(4). The results are then displayed and interpreted to provide decision support for
the user (5). All those phases consider the three life cycle stages: production, use
and end of life.
The first step defines the technical specifications of the product. In this step,
the decision is made whether a single or a multiple product optimization is per-
formed. After this decision, the possible module alternatives, additional measures,
and their availability need to be defined. Possible options are e.g. alternative
materials or material sources, alternative energy supplies for the production or
use stage or physical changes to the product like a change in the geometry. As a
result of this information, the structure of the product and product alternatives is
defined. Section one of figure 4.1 shows the configuration of the product system
and the conversion of the product and the selected measures into a product sys-
tem structure along the entire life cycle. This product system structure includes
alternatives for each choice that represent the available measures. This step is
required to fulfil the requirements R1 to R3. The first step of the framework is
described in more detail in subchapter 4.3.
The second step builds on the results of step one. Considering the defined
alternatives of the product configuration at hand, the life cycle performance is
calculated. The LCA and LCC results for the individual alternatives are calculated
at first as described in Section 2.2 and then recombined according to the defined
product configurations. Section two of figure 4.1 shows how the LCA and LCC
results are assigned to the alternatives of the product system structure. This step
is required to fulfil the requirement R4 and it is described in more detail in
subchapter 4.4.
The third step organizes the configuration-specific information of the individ-
ual product system option into a database where the information is structured
along the hierarchical setup of the product. The data structure is designed in a
way that the individual LCAs and LCCs of each alternative are linked to the
represented life cycle allowing the optimization algorithm to directly take this
information as input data. The results now have to include the individual influ-
ences of the alternative interdependencies and the financial effects of the time
 Raw materials producon Use End of life

B1 D1 G1
+ … C1 E1
A B2 D2 F1 G2 Z1
1

structure
Selecng a basic product, alternave measures for C2 E2
configuraon and modelling the product system structure

Product system
B3 D3 G3

€ GWP B1 GWP GWP D1 GWP GWP GWP G1 GWP

% € € C1 € € E1 € € €
t
% % % % % % %
A B2 D2 F1 G2 Z1
2
Quanfying LCA and LCC and availability while C2 E2

to each node
considering interdependencies and cost development B3 D3 G3

Adding input data

DB DB
B1 D1 G1
C1 E1
€ + GWP € + GWP € + GWP
3 A B2 D2 F1 G2 Z1
4.2 Framework for the Optimization Concept

The individual LCA & LCC results are mapped to the C2 E2

and provision
product structure network to each individual alternave B3 D3 G3

Data management
( )
= ∗ B1 D1 G1
( )
C1 E1
( )
≤ ( )
( )
4 A B2 D2 F1 G2 Z1

LCA & LCC

The opmizaon algorithm based on graph theory is filled C2 E2

opmizaon
with the data to idenfy the most eco-effecve soluon B3 D3 G3

results
The results of the opmizaon are visualized in different

Visualizaon of
forms to support the decision-making process
95

Figure 4.1 Framework for the optimization of eco-effectiveness for modular product systems
96 4 Concept for the Optimization of Eco-effectiveness of Product Systems

span of the use stage that occur due to varying prices for e.g. fuels over time.
The data management is explained in detail in Section 4.4. The third section of
figure 4.1 shows how the individual LCA and LCC results are linked with each
path through the network. This step is required to fulfil the requirements R2 and
R4 and also described in more detail in subchapter 4.4.
The fourth step is the optimization of the input data to identify the ideal
combination of alternatives to result in the product system with the required eco-
effectiveness. The optimization algorithm detects the shortest path of e.g. the
costs through the products systems alternatives while not exceeding an environ-
mental threshold value. This shortest path through the product systems network
represents the recommended alternative combination. The optimization algorithm
is described in Section 4.5 and visualized in section four of figure 4.1. This step
is required to fulfil the requirements R5 to R7 and is described in more detail in
subchapter 4.5.
After the results for the ideal product configuration are calculated, the results
need to be displayed and visualized in order to be interpreted by the user. The
results have to be exported and converted to represent the network structure of
the product system or a graph that shows the ratio of emissions to costs. This is
provided in the fifth step of figure 4.1 and further described in Section 4.6. This
step is required to fulfil the requirement R8. It is also described in more detail in
subchapter 4.6.
The user of the framework requires a user interface that provides relevant
information, display of results and setting options to initiate the optimization
of the product system and to support the decision making during the prod-
uct development. Figure 4.2 shows the minimum requirements for such a user
interface with the four steps of (1) definition of the optimization objective, (2)
configuration of the product system, (3) input data and definition of the financial
time-perspective and (4) calculation and visualization of the results. The detailed
walkthrough through the user interface and the required steps towards the opti-
mization of product systems are explained in an exemplary application cycle in
Section 5.2.

4.2.2 Selection of Graph Theory as Optimization Approach

LCAs and LCCs are descriptive calculations that represent the status quo of the
analyzed product system. However, both LCA and LCC do neither lead to opti-
mized results nor in the ideal product configuration (Cerri et al. 2014). Using the
results of LCA and LCC for the optimization of eco-effectiveness requires the
4.2 Framework for the Optimization Concept 97

I: Definion of the opmizaon objecve

• Definion of the opmizaon objecve (CO 2 or costs)
• Definion of the opmizaon threshold value (e.g., 10 tons of CO 2 or 500 €)

II: Configuraon of the product system

• Configuraon of the product system and its technical specificaons
User interface

• Selecon of single product or mulple product opmizaon and fleet size

III: Input data and definion of the financial me-perspecve

• Calculaon and provision of LCA & LCC input data
• Selecon of relevant price development scenario for raw material
• Selecon of relevant price development scenario for energy and fuels

IV: Calculaon and visualizaon of the results

• Calculaon of ideal product configuraon and numerical export of selected
Data
modules and the resulng environmental impacts and life cycle costs
• Graphical visualizaon of the shortest path through the product system network
and indicaon of selected modules

Figure 4.2 Functionality of the user interface

interpretation of the given results and the application of a suitable optimization

method to the results of the LCA and the LCC. By means of an optimization
approach that combines both results, the ideal product system configuration for a
desired eco-effectiveness can be identified.
As the goal of this approach is the calculation a most eco-effective product
system with the highest eco-efficiency, the optimization of only one target value
(GHG emissions or costs) is required. The algorithm is either supposed to find
the most cost-efficient way to reach the target value of the environmental goal or
to identify the highest possible environmental benefit for a limited budget. From
all possible product system configurations, the ideal solution has to be found.
A special form of modelling problems is offered by the principles of graph
theory. Using the graph theory, many different problem types can be modelled
and solved by suitable algorithms. The advantages of the graph theory, such as
flexible graph design, avoiding of unwanted connections between vertices and
visualization of the results, offer a wide range of possible use cases. The choice
of the graph theory is also based on the broad availability of adaptable algorithms
and the ability to create solutions in a polynomial runtime. As shown in the
previous subchapters, the product system’s architecture can be modelled as a
network considering the possibly occurring interdependencies. Those networks
can directly be applied to a graph searching algorithm. The modelling of networks
and the optimization using the graph theory allows a clear problem structure and
clearly defined module combination options for the algorithm to choose from.
98 4 Concept for the Optimization of Eco-effectiveness of Product Systems

By linking multiple nodes or vertices (V ) with (weighted) edges or arrows

(E), a graph (G) is created that defines and visualizes complex relations between
the vertices. Every element of E has exactly one assigned pair of vertices i and
j coming from V . If a graph consists of edges that each connect two vertices
in both directions, it is an undirected graph. If the edges link two vertices with
a defined direction, the edges are called arrows, making the undirected graph a
directed graph. If the arrows or edges of a graph are assessed with an individual
quantification, the graph is called a weighted graph. (Domschke, Drexl, Klein &
Scholl 2015 p. 72 ff.) The weight of an arrow or an edge can e.g. be the financial
costs of linking two nodes. An induced subgraph (H) of the original graph (G)
is a graph H where the vertices (V (H)) and the edges (E(H)) all are elements of
(V (G)) and (E(G)) (Korte & Vygen 2018, p. 16). Figure 4.3 shows an undirected
graph (G) and an induced subgraph of G(H).

Undirected graph (G) Undirected & induced subgraph (H)

1 3 5 1 5

2 4 6 2 4 6

Figure 4.3 Undirected graph and undirected and induced subgraph (based on Domschke
et al. 2015, p. 72 ff.)

For a given directed or undirected graph it is possible to calculate the shortest

paths between two vertices with the established algorithms by e.g. Dijkstra or
Moore-Bellman-Ford. The algorithms find the shortest path of connected edges
or arrows from a defined starting node s to a desired target node v. . (Korte &
Vygen 2018, p. 168ff) Figure 4.4 shows a weighted and directed graph (G) and
the path from node 1 to node 6 with the least total weight on all arrows (right
network). This path covers all nodes and results in a total weight of 110, hence,
being the “shortest” path compared to all other possible routes. The algorithm
of Dijkstra runs with a time complexity of O(m + n log n) and is solvable in a
polynomial time (Korte & Vygen 2018, p. 170).
A graph (G) is called a network if two nodes of a directed graph are defined
as a source (s) and a sink (t) and each arrow (in networks also called arcs) has
a limited capacity (c). A flow through this network begins at the source (s) and
follows the arcs to the sink (t) without exceeding the given capacity of each
4.2 Framework for the Optimization Concept 99

Weighted & directed Graph (G) Weighted & directed graph (G)
75 75
1 3 5 1 3 5
30 15 10 20 35 30 15 10 20 35

2 35 4 6 2 35 4 6

Figure 4.4 Weighted and directed graph and the shortest path (based on Domschke et al.
2015, p. 72 ff.)

arc. (Korte & Vygen 2018, p. 185 ff.) Building on the definition of networks
in graph theory, network flow problems of different types can be modelled and
solved. The Minimum-Cost Flow Problem (MCFP) is one prominent example.
The MCFP describes a problem in which the flow through a network needs to
be identified that goes from source (s) to sink (t) with the least financial costs
without exceeding the arcs capacities. The costs are expressed as the weights
of each arc. (Korte & Vygen 2018, p. 227 ff.) Figure 4.5 shows an exemplary
network for a MCFP.

Network N = (G, c, s, t)
(35/100)
Costs = 35
Capacity = 100 (65/100) (65/100)
1 3 5 (45/100)
(35/60) (25/70)
S t
(40/40) (45/30)

(20/80) (35/30)
2 (25/75) 4 (50/85) 6
Figure 4.5 Network for minimum-cost flow problem

The mathematical notation of the MCFP for a network like the one shown in
figure 4.5 is shown in equation (4.1) and (4.2). The flow through the network is
represented by f , the edges of the graph (G) are represented by e (first edge) and
E (last edge). The maximum capacity of an edge is represented by c.
100 4 Concept for the Optimization of Eco-effectiveness of Product Systems


E(G)
Minimi ze costs( f ) = f (e) ∗ costs(e) (4.1)
e(G)


E(G)
Subjectto f (e) ≤ c(e) (4.2)
e(G)

While equation (4.1) is the objective function of minimizing the total costs for
the network flow, equation (4.2) limits the individual flows for each edge to its
maximum capacity. For the MCFP, there are different algorithms (e.g. network-
simplex algorithm) that enable the calculation of a solution in polynomial time
(Korte & Vygen 2018, p. 245 ff.).
Given defined arcs between the nodes, the developer defines which con-
nections and thus combinations are possible. Illogical or technically unfeasible
module combinations (e.g. two modules from the same category) can be for-
bidden by leaving out such arcs within the network. Given the defined module
hierarchy, it can also be avoided that certain node choices are skipped during the
optimization. The structure of the network forces the algorithm to make exactly
one choice per node option on its path, avoiding shortened or incomplete product
systems that result in more cost-efficient solutions or solutions with less GHG
emissions (e.g. leaving out the engine of a vehicle to reduce GHG emissions and
costs). Given the target of optimizing the cost-efficiency for an absolute emission
limit of GHG, the optimization regarding a single objective is sufficient. There-
fore, a multi-criteria approach that requires weighting factors between multiple
objectives is not required for the optimization.
Due to the concept of modelling flows through networks, not only single
product units can be optimized, but also product fleets. An additional restriction
lets a defined number of units enter the network and forces the algorithm to have
this number of units also exiting the network. This creates a fixed flow of multiple
units through the network. In this case, the optimization goal is to find the ideal
flow for all product units so that the total costs for all product units in total is
minimized. Building on the flow of multiple product units through the network,
limited carrying capacities of arcs within a network can be used to limit the
availability of single modules. Without limiting the capacity of available modules
by limiting the capacity of the arcs that lead to these modules, the algorithm is
free to use as many modules of every type as it requires for the ideal solution.
This can be applied to modules or options within the product system that cannot
be included in all product unit paths. One example is a limited supply of green
4.2 Framework for the Optimization Concept 101

electricity for the production or the use stage of a product system. Given that,
not every product unit can receive this green electricity supply and the algorithm
needs to identify the product unit configurations that profit the most from the
green energy source and connect the corresponding modules in this way.
The flexibility of the graph design also allows the incorporation of further
nodes that are integrated into the network or into individual paths to represent
possible additional aspects for the respective affected nodes. This can be used e.g.
to consider additional investments that are required before a certain technology
can be implemented. Given an additional node, the cost structure of a module
node does not need to include the investment costs but can remain on the level
of costs per product unit. An example is shown in figure 4.6; the paths that
lead to node D3 first have to flow through the newly added Invest node before
reaching node D3 . In view of a node that represents the required investments for
the following module being incorporated into the network, the optimization of a
single product unit does not include node D3 in the ideal solution due to high
investments for only one product unit. However, if multiple product units enter
the network, the investment that is required to unlock the invest node before node
D3 can possibly be compensated by the savings of now being able to choose node
D3 .

Network

Node Node Node Node Node

A1 B1 C1 D1 E1

Node Node Node Node Node

A2 B2 C2 D2 E2

Node Node Node Node Node

Invest.
A3 B3 C3 D3 E3

Figure 4.6 Integration of additional nodes

Consequently, additional investment nodes act as threshold values that require

a minimum number of product units that are available to include node D3 into
their optimized pathway through the network to make up for the high investment
costs. Considering such an approach of modelling additional costs like invest-
ments that are required for individual modules, aspects like infrastructures or
new factories for new technologies can be included into the model.
102 4 Concept for the Optimization of Eco-effectiveness of Product Systems

Given these advantages, the optimization using the graph theory offers several
options to identify the ideal solution regarding various flexible framing condi-
tions. The identification of the ideal path through the network with the weighted
arcs can be transferred to the search for the shortest path through a network. The
shortest path for e.g. a given emission limit results in the shortest path regard-
ing the costs of the product configuration. When the optimization objective is
switched from emissions to costs, the shortest path for a given financial bud-
get results in the product configuration that offers the minimal path of GHG
emissions.
Due to clearly defined modules and the individual use stage that belongs to
each possible product configuration, it is possible to add further constraints to the
solution space regarding threshold values for selected life cycle stages. For many
product types it is relevant to not only contribute to sustainability by reaching
emission levels along the entire life cycle, but also being below limiting values
e.g. for the production stage or during the use stage due to political regulations.
Figure 4.7 shows an example of a product system with a limit of 100 kg of CO2
eq. over the entire life cycle. The solution with the lowest GWP in this example
is product system A which emits 30 kg of CO2 eq. during the production, 50
kg of CO2 eq. during the use stage and 20 kg of CO2 eq. during the end of life
stage.
The product system A consequently complies with the life cycle limit of maxi-
mum 100 kg of CO2 eq. If an additional limitation, e.g. for the use stage, is added,
the algorithm possibly has to find a new solution which meets both limitations.
If such a new limitation for the use stage is added which allows a maximum of
30 kg of CO2 eq. for this stage, product system A is not a valid solution any-
more. In such a case, the algorithm has to switch to product system B, which is
more expensive in this example but complies with both emission limitations at
the same time.
A focus on only the use stage, however, leads to missing absolute sustainability
goals, as other life cycle stages still emit GHG. This makes it a requirement to
meet limits for selected life cycle stages (e.g. use stage) and limits for the entire
life cycle at the same time. Due to the modular design of the network, such
additional emission limits for individual life cycle stages can be integrated into
the solution identification.
Furthermore, the results of the shortest path identification can easily be
visualized in the already modelled product system network. In view of these
characteristics, an optimization algorithm based on graph theory offers a solution
approach that covers the stated requirements to an approach that were introduced
in Section 3.1 and Section 4.1.
4.3 Modelling the Product System’s Structure in a Network 103

0 kg CO2 eq. 25 kg CO2 eq. 50 kg CO2 eq. 75 kg CO2 eq. 100 kg CO2 eq. 125 kg CO2 eq.

Product end of
production use
system A life

Product
production use end of life
system B

Use phase limit

of 30 kg CO2 eq.

Life cycle limit of 100 kg CO2 eq.

Figure 4.7 Multiple limitations for different life cycle stages

4.3 Modelling the Product System’s Structure

in a Network

This subchapter describes step one of the framework depicted in figure 4.1.
Here, it is shown how the product system is modelled as a network consider-
ing interdependencies between components or modules. As the incorporation of
interdependencies leads to a strong growth of the network, reduction strategies
are introduced to reduce the demand of required nodes to model the product sys-
tem. The procedure of structuring and mapping the product system’s architecture
into a network that represents the possible combinations of modules and compo-
nents with respect to the technical limitations of certain combinations forms the
basis for the data generation.

4.3.1 Transformation of Modular Product Systems

into Networks

Figure 2.3 in Section 2.1 introduced the architecture of modular product units
and figure 2.6 showed the evolution to fleets of modular product systems. These
104 4 Concept for the Optimization of Eco-effectiveness of Product Systems

depictions, based on the life cycle perspective and the product architecture, are
suitable to explain the general structure of a product system, but are not applicable
to graph theory algorithms. The foundation for the optimization framework is
built by the possible product system configurations that form the options and
thus the search space from which the algorithm combines the best options to the
final product system. These product system configurations in return are defined by
the available components and modules that can be used to assemble the product.
The product system architecture and the adjunct life cycle stages need to be
transferred into a network structure to be able to apply an optimization algorithm
to the product system.
Figure 4.8 shows how a simple modular product unit is transformed into a
network structure. The product’s architecture structure is defined with a directed
graph, where each alternative for the modules is represented with an individual
node. The exemplary modular product structure in figure 4.8 is built from four
different modules (module A, module B, module C and module D). Each mod-
ule forms its own “module category” or “module family” that provides a unique
function to the product. While module A of this example only offers one alterna-
tive within its module category, module category B offers two alternatives B1 and
B2 with the same functionality. Module category C offers three different alter-
natives to choose from (C1 , C2 and C3 ) and module category D again provides
two alternatives (D1 and D2 ). Given such a network structure for the product, the
connecting arrows are weighted with e.g. costs or GWP that are associated with
each module. This set-up allows shortest path algorithms to be applied.
The network has to be followed from module A to module D along the direc-
tion of the connecting arrows to build an entire product unit. For every module
category, only one, but also always a minimum of one alternative, has to be
chosen. The grey highlighted modules in figure 4.8 show one possible way of
module choices along the network to result in one product unit. In this work, the
terms “module alternatives”, “module variants” and “module options” are used
synonymously to describe alternatives for modules with the same functionality
that come from the same module category.
Further nodes and node categories need to be included to expand this network
of a product unit to an entire product system including the life cycle perspec-
tive. These additional nodes represent alternative choices for adjunct processes
regarding the production stage or the end of life stage. Figure 4.9 shows the evo-
lution of the network in figure 4.8 into a network that represents not only the
product unit, but also exemplary life cycle stages with their own node categories
and alternatives. The product unit is embedded in adjunct choices for supportive
4.3 Modelling the Product System’s Structure in a Network 105

functions. Such a network that represents the modular product system will fur-
ther be referred to as a “product system structure network”. This network adds
complexity by including additional life cycle stages via new node categories and
by including the product architecture level of components into the modules. The
lowest architecture level of parts is not included to reduce complexity for this
work and for the visualization of the networks.

Modular product unit

module C1

module B1 module D1

module A module C2

module B2 module D2

module C3

module category module category module category module category

A B C D

Figure 4.8 Modular product unit as a network (based on Jose & Tollenaere 2005, p. 374)

Figure 4.9 shows an exemplary product system structure consisting of two dif-
ferent modules (module A and module B). These two modules are complemented
by an upstream category for the production stage (P1 ) and a downstream category
for the end of life processes (EoL1 ). Within the product unit, both module cate-
gories A and B of the product unit consist of two alternative modules (A1 , A2 and
B1 , B2 ) which enables alternative paths for the product configuration resulting in
four different options for the product system design (P1 –A1 –B1 –EoL1 , P1 –A1 –
B2 – EoL1 , P1 –A2 –B1 –EoL1 and P1 –A2 –B2 –EoL1 ). On a deeper level into the
product unit structure, module A and module B consist of three components (A,
B and C). Each component has a total of three alternatives available (A1 , A2 &
A3 | B1 , B2 & B3 | C1 , C2 & C3 ). For each module this offers nine different com-
ponents that are structured in three different component categories. From every
available component category, one option has to be selected.
106 4 Concept for the Optimization of Eco-effectiveness of Product Systems

On the level of components, the options cannot be combined freely. The com-
ponents A1 and A2 can be combined in all manners with B1 and B2 and these
components again with C1 and C2 , which leads to eight alternative combinations
between these six components. The components A3 , B3 , and C3, in contrast,
form an exception as these components can only be combined as a trio due to
exemplary technical limitation that excludes them from the other options. The
arrows in the network show only one option for each of these components to
represent these limited combinatorial options. Each entire module can be built in
nine different configurations. The production stage and the end of life stage also
offer nine alternatives with the same combinatorial logic as the product unit’s
modules. For the production stage and end of life stage, three alternatives for the
location (L1 to L3 ), three alternatives for the electricity mix (E1 to E3 ) and three
alternatives for different machines (M1 to M3 ) can be used.
Transferred to the automotive industry, the module category A can e.g. be the
choice of powertrain, with module A1 being a combustion engine-based pow-
ertrain and module A2 an electric powertrain with a high-voltage battery for
electricity storage. Module category B can represent different alternatives for the
car body, with module B1 being a conventional, steel-based car body and module
B2 a lightweight construction with a higher share of aluminum.
The four product system categories with each two alternatives for module A
and module B of the product unit offer a total of 54 alternatives from which
twelve alternatives (three alternatives within the four categories) are part of the
final product system. The grey highlighted nodes in figure 4.9 show one possible
way of the option selection within the product system structure network. Provid-
ing nine options for the configuration of each the production and end of life stage
and also for all four module alternatives, the network shown in figure 4.9 offers
a total of 26, 244 different options of combinations to create a complete product
system.

4.3.2 Product System Networks Including Interdependencies

Based on the principle of modular LCAs and modular LCCs, individual assess-
ments for the 54 alternatives of figure 4.9 are sufficient to analyze all those 26,244
possibilities as the rest of the required information for the entire product system
can be derived from the 54 initial results. This approach meets its limit when
interdependencies between modules or components influence the performance, so
that static LCAs or LCCs for modules do not apply to every variation of module
combination. As the LCA or LCC of a single component or module can possibly
 modular product system structure network
modular vehicle
module A1: combustion engine module B1: conv. car body
Comp. Comp. Comp. Comp. Comp. Comp.
CA1 CB1 CC1 CA1 CB1 CC1
Comp. Comp. Comp. Comp. Comp. Comp.
CA2 CB2 CC2 CA2 CB2 CC2
production stage P1 end of life stage EoL1
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L1 E1 M1 CA3 CB3 CC3 CA3 CB3 CC3 L1 E1 M1
Loc. Electr. Mach. Loc. Electr. Mach.
L2 E2 M2 L2 E2 M2
module A2: electric powertrain module B2: lightweight car body
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L3 E3 M3 CA1 CB1 CC1 CA1 CB1 CC1 L3 E3 M3

Comp. Comp. Comp. Comp. Comp. Comp.

CA2 CB2 CC2 CA2 CB2 CC2
Comp. Comp. Comp. Comp. Comp. Comp.
CA3 CB3 CC3 CA3 CB3 CC3
4.3 Modelling the Product System’s Structure in a Network

production category module category A module category B end of life category

Figure 4.9 Product system structure network of a modular product system

107
108 4 Concept for the Optimization of Eco-effectiveness of Product Systems

change if the module is combined with another alternative, the sum of the static
LCA values and LCC values results can differ from the value that arises when
including interdependencies. The interdependencies need to be modelled directly
into the product system’s structure network to overcome the limitations of regu-
lar modular assessments and to include the effects of interdependencies into the
calculation.

Modelling general interdependencies into networks

Interdependencies between modules can change the performance of other modules
and thus the specifications of the whole product system. The way how the interde-
pendencies affect the modules of a product system can be different for every case.
Some interactions may only relate to one module, some may influence all involved
modules and sometimes it is possible that a change of a module does not cause any
interdependencies at all. In a worst-case scenario, interdependencies occur between
every module of the product system.
If a product unit consists of only two modules (represented by node A and node
B in figure 4.10) and each module offers two alternatives (represented by node A1 ,
node A2 , node B1 , node B2 ), the structure network of this product system looks like
Network A in figure 4.10. The exemplary selected design alternative for this product
is node A2 combined with node B1 , as highlighted in grey. If interdependencies
exist between all design options and the performance of the whole product system
depends on the exact combination of the modules, the results for the LCC and LCA
of e.g. node A1 differs when it is combined with either node B1 or node B2 . For
node A1, two different states exist, one for each possible combination it is part of.
This logic also applies to node A2 , node B1 and node B2 .
Network A II in figure 4.10 shows how a network is modelled that incorporates
the different states of each node. For every node, two different “subnodes” exist.
Each subnode is connected to only one specific node of the other class to indicate the
combination partner that it is exclusively meant for. The required number of nodes
in the network increases from four to eight to fully represent the interdependencies
between nodes, while the number of connecting arrows does not change with four
remaining edges. Also, the number of possible combinations remains at four options.
The selected combination of nodes in Network A II is therefore clearly defined given
node A2–1 and node B1–2 .
Transferring the principle of incorporating interdependencies into the exemplary
network design of Network B leads to an enormous growth of the number of nodes,
assuming interdependencies are occurring within every configuration alternative and
for every node. Network B offers a total of 729 different combination possibilities.
One alternative from each node category (A to F) has to be selected to create a
4.3 Modelling the Product System’s Structure in a Network 109

complete product system. Consequently, node A1 needs to be connected with five

more nodes. As each of the five remaining category offers three choices, node A1 can
be matched with 243 different following configurations. This leads to a required split
up of node A1 into subnodes ranging from node A1–1 to node A1–243 and following
that logic to the same split up for all other 17 nodes of this network. The number
of required nodes thus increases from 18 nodes to 4, 374 nodes. Based on these
4,374 nodes, the complete network can be represented with respect to possible
interdependencies between all nodes.

Network A II

Node Node
A11 B11
Subnodes
Network A for Node
Node Node
B1
Node Node A12 B12
A1 B1

Node Node
A2 B2 Node Node
A21 B21

Node category Node Node

A A22 B22

Network B

Node Node Node Node Node Node

A1 B1 C1 D1 E1 F1

Node Node Node Node Node Node

A2 B2 C2 D2 E2 F2

Node Node Node Node Node Node

A3 B3 C3 D3 E3 F3

Figure 4.10 Structure networks with interdependencies

The strong growth of required nodes to represent the interdependencies leads

to an increase in data demand as every node has to be assessed individually to
be used in a modular analysis and optimization. Without interdependencies, the
110 4 Concept for the Optimization of Eco-effectiveness of Product Systems

product system of Network B and all 729 options can be assessed by modular LCA
and LCC by assessing the 18 different nodes individually and thereby reducing the
effort to only 2.5% compared to assessing all 729 options directly. Considering the
expanded network though, the effort resulting from a modular approach rises to
600% as then 4,374 instead of 729 assessments are required. Instead of simplifying
the assessments, the modular approach including interdependencies makes it more
complicated as the number of nodes surpasses the number of possible initial node
combinations. The product system structure in figure 4.9 quickly becomes too big
to depict if all components had possible interdependencies with each other. This
leads to a data demand that is too large to be calculated.
Strategies are needed that help to reduce the effort of modular LCC and LCA
and end up below the threshold of the possible module combinations (in this case
below 729). Such strategies help to still be able to use the modular product approach
for the assessment of the life cycle performance and for the product optimizations.
(Gabrisch, Cerdas & Herrmann 2019)

4.3.3 Network Reduction Strategies for Interdependency

Modelling

The reduction of the effort of data calculation, network modelling and optimiza-
tion can be achieved by applying network reduction strategies to the model of the
product structure to reduce the number of nodes. The number of required unique
nodes to represent an interdependency-sensitive product structure network is cal-
culated using equation 4.3. The variable N represents the number of required
individual nodes, y represents the number of node categories and nx the number
of alternatives for each node category. The number of required nodes depends
on the number of alternatives per node category and on the number of node
categories.
 
N = y ∗ n1 ∗ n2 ∗ . . . ∗ n y (4.3)

The number of node categories defines the “length” of the product system’s struc-
ture. The length of the product structure network does not influence the solution
quality as it simply defines the product system’s required functionality and design.
The number of alternative options per node category, however, is a generally
4.3 Modelling the Product System’s Structure in a Network 111

desired state, as more options lead to a bigger search space for the optimization
and possibly to better results.

Reduction via network partitioning

As shown in equation 4.3, a “long” structure network with many connected node
categories increases the need for individual nodes, represented by the factor y.
Therefore, shortening the length of the structure network directly reduces the node
demand. A shortening of the structure network can be achieved by partitioning
the complete structure network into “subnetworks” or “substructures”. With such
a split, two separated subnetworks are created, both having fewer node categories
to calculate with. This reduced number of node categories then reduces the number
of required individual nodes to model the interdependencies. The two subnetworks
are then connected with a single connecting node serving as a bridge between the
subnetworks.
However, such a network partition can only be done when it can be guaranteed
that there are no interdependencies occurring between nodes of different subnet-
works. Interdependencies are only considered within each subnetwork, but not
between them. This can be applied when (referring to figure 4.9) e.g. the production
stage of a vehicle has interdependencies between the choice of machines and the
selected energy mix, but none of the production stage nodes has interdependencies
with the following use stage. Then, this network can be partitioned into a subnetwork
for the production stage with its internal interdependencies and a subnetwork for
the use stage with again its own internal interdependencies. However, now no nodes
are required to model the non-existing interdependencies between the production
stage and the use stage.
A theoretical split of the product structure network of Network B in figure 4.10
into two separate chains with each having three connected node categories (see
figure 4.11) leads now to only 8 1 individual nodes both for structure I and structure
II (Gabrisch, Cerdas & Herrmann 2019). In this scenario, the alternatives of node
category A, B and C can now only have interdependencies with each other, but they
do not have a direct connection to the nodes of category D, E and F anymore. This
means that interdependencies between e.g. node B1 and D1 now are not represented
anymore and therefore no nodes are necessary to represent their relation. In sum,
a separated structure network of Network B only needs 162 individual nodes to
represent the network with interdependencies, reaching a number below the 729
combination options shown before.
112 4 Concept for the Optimization of Eco-effectiveness of Product Systems

Network B II

Node Node Node Node Node Node

A1 B1 C1 D1 E1 F1

Node Node Node Node Node Node

A2 B2 C2 D2 E2 F2

Node Node Node Node Node Node

A3 B3 C3 D3 E3 F3

Structure I Structure II

Figure 4.11 Network B with a separated structure network

Reduction using network adaption to include a correction factor

The introduced reduction of required nodes using network separation only works
when interdependencies between several node categories (along the separation) can
definitely be excluded. If interdependencies cannot be neglected or if a subnetwork
itself is still large, a reduction strategy to decrease the number of nodes is required
also within these subnetworks.
As shown by equation 4.3, the demand of nodes to model all interdependencies
grows quickly if the assumption is made that the interactions occur between all
nodes and every node is affected in its own performance by these interactions. Each
node requires multiple subnodes for every interaction to represent those interdepen-
dencies in the network. Here lies a huge reduction potential. Instead of considering
the individual interdependencies between each node pair, the information of the
interdependency between all connected nodes can be concentrated and bundled into
one single additional node. This newly added node as a newly added node category
includes the information for all interactions that occur in a specific product system
configuration and adds a “correction factor” at the end of each network path. This
is feasible as only the interdependencies of all nodes coming together are relevant
and not the interdependencies between two nodes within a product system because
the target is to build an entire product system and not a fraction of it. In figure 4.12,
this alternative modelling of the network is shown.
For a network with three node categories and two alternatives per node like
Network C in figure 4.12, the expanded network requires 24 nodes to depict all
possible interdependencies (see Network C II). Network C III shows the network
that represents the interdependencies with an additional node category N. This type
4.3 Modelling the Product System’s Structure in a Network 113

of network is designed in a way that each node category does not contain all possi-
ble interaction information anymore. Instead, only the individual path through the
network is modelled. While a node can have multiple outgoing arrows to the follow-
ing node category, the incoming arrows always have a unique and clearly defined
source. In this model, the network size grows with each category.
The first categories in Network C III are relatively small compared to Network
C II. The full set of alternative subnodes for interdependencies only occurs in the
last node category. An additional fourth category is added at the end of the network
to make up for the lost information along the reduced node categories. That new
category holds an individual node for each subnode of the previous category (node
C), as each subnode of category C stands for a unique pathway through the network.
This new network category (N) contains the summarized information of the interde-
pendencies of all nodes within this product system configuration. The selected path
of A2 , B1 and C1 results in N2 , where the interdependencies of this configuration
are stored. The required number of nodes to create the new network is reduced, as
now 22 nodes (Network C III) instead of 24 nodes (Network C II) are needed.
This is a reduction of required nodes by 8.33% for this example. For larger
networks, the reduction quickly becomes more effective. If Network C would consist
of ten node categories (A to J) with each category having two node alternatives,
1, 024 possible combinations existed. Based on equation 4.3, Network C II would
need 10,240 nodes to represent all possible interdependencies. Network C III only
needs 3,070 nodes to provide the same information as Network C II, resulting in a
reduction of 70% of the required nodes. If Network C consisted of ten node categories
with three alternatives each (e.g. A1 to A3 ), Network C III would reduce the number
of nodes by 75% in comparison to Network C II. The larger the network regarding
node categories and number of node alternatives per category becomes, the more
effective the node reduction strategy of the additional correction node becomes as
well. While the number of nodes for Network C II increases exponentially, the
number of nodes for Network C III grows in a linear way. Introducing a correction
factor N as an additional category at the end of the network helps to reduce the
number of required nodes compared to a full expansion of the network for all nodes
and interactions. Still, the number of nodes (and thus potentially the data demand)
is higher than the initial number of possible combinations. However, in this type
of network modelling, the number of nodes does not represent the required data as
the individual subnodes for the node categories do not represent individual data,
but only enable the tracking of the unique paths through the network. Hence, e.g.
the subnodes C2–1 to C2–4 are four individual subnodes, but they all contain the
same information or data, making only one calculation necessary to provide the
information for four subnodes.
 114
4

Network C II Network C III

Node Node Node Node

N1
A1-1 B1-1 C1-1 C1-1
Node Node Node Node
N2
A1-2 B1-2 C1-2 C1-2
Node Node Node Node Node
B1-1 N3
A1-3 B1-3 C1-3 C1-3
Network C Node Node Node Node Node
Node N4
Node A1-4 B1-4 C1-4 B1-2 C1-4
Node Node A1
A1 B1 C1

Node Node Node Node Node Node Node Node Node

A2 N5
A2 B2 C2 A2-1 B2-1 C2-1 B2-1 C2-1
Node Node Node Node
Node N6
A2-2 B2-2 C2-2 C2-2
B2-2
Node Node Node Node
N7
A2-3 B2-3 C2-3 C2-3

Node Node Node Node

N8
A2-4 B2-4 C2-4 C2-4

Figure 4.12 Network structure with correction factor

Concept for the Optimization of Eco-effectiveness of Product Systems
4.3 Modelling the Product System’s Structure in a Network 115

4.3.4 Strategy Adaption to Reduce the Data Demand of LCA

Values and LCC Values

As shown in the previous section, the implementation of possible interdependen-

cies into the network leads to a strong increase in required nodes and input data.
Therefore, reduction strategies have been introduced that show how the network
size is reduced down to a manageable size. These network-based strategies need
to be applied to the product system’s structure network and the modelling of
LCC and LCA data to transfer the node reduction into a data reduction. In this
subchapter, it is shown that the network reductions also result in a reduction of
data demand for individual LCA and LCC assessments. The target is modelling
the product system structure on component and module level that includes the
interdependencies, but still enables a modular LCA and LCC assessment and a
modular optimization.

Reduction using network partitioning

The reduction strategy of network partitioning reduces the number of nodes by
splitting up a network into two subnetworks. By this separation, the two subnet-
works become independent from each other. Interactions between the components
of the separated networks are no longer considered and therefore do not need to
be modelled with specific nodes. Partitioning networks can only be done between
components or modules where it is known that no interdependencies occur to avoid
losing information of interdependencies. Building on the definitions of modularity
for product systems as shown e.g. by Mutingi et al. (2017) (see Section 3.2.2), ideally
the liaisons between the components within a module are maximized and the liaisons
between modules minimized (Mutingi et al. 2017, p. 473). Based on the assumption
that modular products show no interdependencies between the components of one
module and the components of another module, the network partitioning can be
applied. For structures that show a high modularity, a separation of the whole graph
into subgraphs is allowed at the borders of individual modules. However, interde-
pendencies within the components of the same module and also interdependencies
between different modules are still possible. Still, the network partitioning along
the modules reduces the required data for LCA and LCC. It depends on the number
of modules within the structure network and the number of components per module
to which extend the data demand is reduced.

Reduction by network adaption to include a correction factor

The application of the network size reduction strategy that provides a correction
factor for each path through the structure network to a real product system is more
116 4 Concept for the Optimization of Eco-effectiveness of Product Systems

complex. The implementation of a correction factor is built on the idea that all inter-
dependency influences are summarized into one overarching node that represents
the entire network path it is part of. The product to be analyzed has to behave in this
way as well in order to transfer this form of modelling to the life cycle performance
of a product system. This means that all interactions and interdependencies between
the components of a module or the modules of a product system are centralized at
one point to be quantified and included to the network structure. A solution for the
transfer of the correction factor to a product system is the separation of the prod-
uct’s production and end of life stage from the use stage of the product. Figure 4.13
depicts this separation.
The foundation to the incorporation of the correction factor is that the interde-
pendencies that are caused by the changed product configurations of module or
component alternatives all interact during the use stage. This means that all changes
to e.g. the weight or energy provision of the product system (e.g. a vehicle) are
effective during the use stage. During this use stage, all interactions are active at the
same time to provide the desired functionality of the product. Before entering the
use stage, it is not possible to assess the influence that a reduced weight has on e.g.
the choice of the engine or other modules. Therefore, only the use stage is suitable to
detect and quantify the occurring interdependencies of all modules. The separation
between the production on one side and the use stage on the other side is performed
since in order to enter the use stage, the entire product system has to be designed
and manufactured completely. This means that for every component and module
category a choice has been made. Changes to this product configuration are not pos-
sible anymore and the individual use stage of this defined configuration begins with
all information regarding this product system configuration being available (e.g. the
total weight or the air resistance coefficient). This information helps to assess the
financial and environmental impacts of the use stage specifically for each individual
product configuration. The separated use stage in figure 4.13 has a unique subnode
for every possible product configuration, just as in figure 4.12 every possible path
through the network has an own node N which provides the correction factor.
This allows the adaption of the use stage module in the LCA and LCC assessment
as the correction factor. The correction factor serves as a corrective dimension that
balances out the deviation of the added static module results from the true values that
belong to the specific combination. For a product system this means that within the
network for the production and end of life stage, the information of the individual
modules is added up and matched with the corresponding use stage. In this way, the
interdependencies do not have to be considered during the production and end of
life stage, but only during the use stage. Incorporating the interdependencies into
the use stage requires a limitation of these interdependencies to the use stage only.
 modular product system structure network
network I network II

modular vehicle
use 1
module A1: combustion engine module B1: conv. car body
Comp. Comp. Comp. Comp. Comp. Comp.
CA1 CB1 CC1 CA1 CB1 CC1
use 2
Comp. Comp. Comp. Comp. Comp. Comp.
CA2 CB2 CC2 CA2 CB2 CC2
production stage P1 end of life stage EoL1
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach. use 3
L1 E1 M1 CA3 CB3 CC3 CA3 CB3 CC3 L1 E1 M1
Loc. Electr. Mach. Loc. Electr. Mach.
L2 E2 M2
module A2: electric powertrain module B2: lightweight car body L2 E2 M2 + use 4
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L3 E3 M3 CA1 CB1 CC1 CA1 CB1 CC1 L3 E3 M3
…
Comp. Comp. Comp. Comp. Comp. Comp.
CA2 CB2 CC2 CA2 CB2 CC2
Comp. Comp. Comp. Comp. Comp. Comp.
CA3 CB3 CC3 CA3 CB3 CC3
use 26,244
4.3 Modelling the Product System’s Structure in a Network

production category module category A module category B end of life category use phase category

Figure 4.13 Separation of use stage from product system network

117
118 4 Concept for the Optimization of Eco-effectiveness of Product Systems

The selection of an alternative module must not influence the production and end
of life stage of another module.
Referring to an automotive product system, this means that changing the material
of the car body from steel to aluminum influences the use stage of the final vehicle as
the weight is reduced. However, the production and end of life treatment of another
module, e.g. the engine, is affected by the change in the car body. Therefore, both
alternatives are combined with the same engine. Consequently, the financial and
environmental impacts of the engine do not change in Network I of figure 4.13. The
impact of the changed weight only occurs in the specific use stage for both possible
engine and car body combinations (engine + steel, engine + aluminum).
By adding more module categories to the product system, the direct interdepen-
dencies between car body and engine are now not relevant anymore. The relevant
interactions of all modules (e.g. engine, car body, tire and further modules) together
define the interactions of all modules of the product system and are represented
by the excluded use stage. For a product system that consists of ten modules, the
interdependencies between the first and the second module are not relevant. Only
the interdependencies that occur when all modules are combined at the same time
need to be considered. The change of one module of the product system does not
need to be evaluated regarding the new interdependencies between the new module
and every other module individually, but only for the whole product system in total.
A split up of modules into subnodes for every module and for every possible inter-
dependency with every other module is not required. The only required information
regarding the interdependencies is needed for the combined interdependencies of all
modules. This information can be combined into one node—the node that contains
the correction factor, which is the use stage for vehicles.
As the required module information for Network I in figure 4.13 are now all static
and not interaction-specific, the information for Network I can be provided with
regular LCA and LCC assessments as described in Section 2.2. The dynamic infor-
mation of the individual influence of each unique product configuration therefore
needs to be provided by the data of the use stage in the network.
The financial and environmental burdens of the use stage of a vehicle depend
on the consumed energy, fuels, and other operating materials during the time of
operation. For the calculation of the individual consumption of the required energy
and material input for the use stage, the physical properties of each configuration are
the defining parameters. Considering the weight, geometry and energy efficiency of
the individual modules, each product configuration has individual specifications that
influence its energy and material consumption for the use stage of a certain product
system. The individual energy and material consumption rates are multiplied by
fixed factors regarding financial costs and environmental impacts. For example, the
4.3 Modelling the Product System’s Structure in a Network 119

demand of electric energy of the product in the entire use stage is multiplied by the
price for one kWh of electric energy and an environmental impact factor. Following
this approach, each individual product system’s use stage is assessed in terms of its
financial and environmental impact by matching its individual material and energy
demand with fixed factors per consumed unit.
An important difference that occurs when transferring the idea of the correction
node from the network modelling perspective to a product system network affects
the number of required nodes and input data. While Network C III in figure 4.14
requires 22 nodes to model the network to allow the tracking of the individual
paths, the subnodes for each module all contain the same static information. As the
interdependencies are all bundled into the use stage, the module configuration does
not need a split up for modelling interactions, only for pathfinding. Regarding the
input data for LCA and LCC, e.g. the nodes of module C1–1 to module C1–4 all
contain the same information. Therefore, the number of nodes does not represent
the number of required input data. Figure 4.14 visualizes the difference between the
node demand (left) and the data demand (right).
For Network C IV, the data demand is calculated by the number of possible
combinations added with the number of nodes to model the static product structure
network. In Network C IV, the possible number of combinations (eight) and the
number of nodes (six) result in 14 required inputs. For larger networks, the difference
between possible combinations and the data demand decreases rapidly. Already for
slightly larger networks, this difference becomes relatively small. For modelling
a network with ten module categories and three options per category considering
the interdependencies, the 59, 049 possible combinations require 59,079 inputs or
0.05% more effort than calculating all options directly. For even larger network this
difference becomes insignificantly small.
Applying the two introduced network reduction strategies to a product system’s
structure network shows that the reduced number of required nodes are converted
into a reduced need for input data regarding the LCA and LCC of the individual
components, modules, and the entire product system. As stated in Section 4.3.1,
implementing the interdependencies between all components and modules into the
network of figure 4.9 results in a network that is too large to model, to provide
input data for and too complex to optimize it. However, by applying the introduced
network reduction strategies, it becomes possible to model this network.
Figure 4.15 shows the application of the reduction strategies to the network in
figure 4.9. Based on the first reduction strategy (network partitioning) and its transfer
to the modularity of product systems, the interdependencies between components
of different modules can be neglected. This allows a focus on interdependencies on
a component level only within each individual module, but not between different
120 4 Concept for the Optimization of Eco-effectiveness of Product Systems

Network C III Network C IV

Node
N1 N1
C1-1

Node
N2 N2
C1-2
Node Node
B1-1 N3 N3
C1-3

Node Node Node Node Node

Node N4 N4
B1-2 C1-4 A1 B1 C1
A1

Node Node Node Node Node Node

A2 N5 N5
B2-1 C2-1 A2 B2 C2
Node
Node N6 N6
C2-2
B2-2
Node
N7 N7
C2-3

Node
N8 N8
C2-4

Figure 4.14 Difference between node demand and data demand

modules. This reduces the interdependency possibilities enormously. The interde-

pendencies between the components within a single module on the other hand are
modelled in line with the network reduction strategy of introducing the correction
factor. Interactions are also possible on the architecture hierarchy level of modules.
These interdependencies are modelled in the same way following the principle of
the excluded use stage information. This results in a unique node of the additional
node category for each possible path.
Following both reduction strategies, it is possible to represent the structure
network including possible interdependencies in a modular way enabling the appli-
cation of modular LCA and LCC and an optimization based on the graph theory. The
number of required nodes to model the network of figure 4.9 lies with 290 nodes far
below the number of possible combinations (26,244). The required input data with
106 data sets of each LCA and LCC is even lower. This number is derived as each
module has nine static LCAs or LCCs for every component and additionally eight
individual use stages for the component combinations. These 17 module specific
data need to be calculated for all six available modules (A, B1 , B2 , C1 , C2 , D).
Finally, the four different use stage configurations for the product system have to
be added.
4.3 Modelling the Product System’s Structure in a Network 121

modular product system structure network

modular vehicle
module A1: combuson engine module B1: conv. car body
Comp. Comp. Comp. Comp. Comp. Comp.
CA1 CB1 CC1 CA1 CB1 CC1
Comp. Comp. Comp. Comp. Comp. Comp.
CA2 CB2 CC2 CA2 CB2 CC2
producon stage P1 end of life stage EoL1
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L1 E1 M1 CA3 CB3 CC3 CA3 CB3 CC3 L1 E1 M1
Loc. Electr. Mach. Loc. Electr. Mach.
L2 E2 M2 L2 E2 M2
module A2: electric powertrain module B2: lightweight car body
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L3 E3 M3 CA1 CB1 CC1 CA1 CB1 CC1 L3 E3 M3

Comp. Comp. Comp. Comp. Comp. Comp.

CA2 CB2 CC2 CA2 CB2 CC2
Comp. Comp. Comp. Comp. Comp. Comp.
CA3 CB3 CC3 CA3 CB3 CC3

producon category module category A module category B end of life category

product system structure network Module B1 Comp.

C1-1
N1 EoL1 Comp.
C1-1
N1

Comp. Comp.
C1-2
N2 C1-2
N2
Comp. Comp. Comp. Comp.
B1-1 C1-3
N3 B1-1 C1-3
N3

Comp. Comp. Comp. Comp.

Comp. N4 Comp. N4
B1-2 C1-4 B1-2 C1-4
A1 A1

Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5
Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5 N1
Comp. Comp.
Comp.
C2-2
N6 Comp.
C2-2
N6
B2-2 B2-2
Comp. Comp.
C2-3
N7 C2-3
N7

Comp. Comp.
C2-4
N8 C2-4
N8

Comp. Comp. Comp. Comp. Comp. Comp.

A3 B3 C3 - A3 B3 C3 -

Module A1 Comp.
C1-1
N1 Module B1 Comp.
C1-1
N1 EoL1 Comp.
C1-1
N1

Comp. Comp. Comp.

C1-2
N2 C1-2
N2 C1-2
N2
Comp. Comp. Comp. Comp. Comp. Comp.
B1-1 C1-3
N3 B1-1 C1-3
N3 B1-1 C1-3
N3

Comp. Comp. Comp. Comp. Comp. Comp.

Comp. N4 Comp. N4 Comp. N4
B1-2 C1-4 B1-2 C1-4 B1-2 C1-4
A1 A1 A1

Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5
Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5
Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5 N2
Comp. Comp. Comp.
Comp.
C2-2
N6 Comp.
C2-2
N6 Comp.
C2-2
N6
P1 Comp.
C1-1
N1
B2-2
Comp.
B2-2
Comp.
B2-2
Comp.
C2-3
N7 C2-3
N7 C2-3
N7
Comp.
C1-2
N2
Comp. Comp. Comp.
C2-4
N8 C2-4
N8 C2-4
N8
Comp. Comp.
B1-1 C1-3
N3

Comp. Comp. Comp. Comp. Comp. Comp. Comp. Comp. Comp. Comp. Comp.
Comp.
A1 B1-2 C1-4
N4
A3 B3 C3 - A3 B3 C3 - A3 B3 C3 -
Comp. Comp. Comp.
A2 B2-1 C2-1
N5 Module A2 Comp.
C1-1
N1 Module B2 Comp.
C1-1
N1 EoL1 Comp.
C1-1
N1
Comp. Comp. Comp. Comp.
Comp.
C2-2
N6 N2 N2 N2
B2-2 C1-2 C1-2 C1-2
Comp. Comp. Comp. Comp. Comp. Comp. Comp.
C2-3
N7 B1-1 N3 B1-1 N3 B1-1 N3
C1-3 C1-3 C1-3
Comp. Comp. Comp. Comp. Comp. Comp. Comp.
C2-4
N8 Comp. N4 Comp. N4 Comp. N4
B1-2 C1-4 B1-2 C1-4 B1-2 C1-4
A1 A1 A1

Comp.
A3
Comp.
B3
Comp.
C3 - Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5
Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5
Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5 N3
Comp. Comp. Comp.
Comp.
C2-2
N6 Comp.
C2-2
N6 Comp.
C2-2
N6
B2-2 B2-2 B2-2
Comp. Comp. Comp.
C2-3
N7 C2-3
N7 C2-3
N7

Comp. Comp. Comp.

C2-4
N8 C2-4
N8 C2-4
N8

Comp. Comp. Comp. Comp. Comp. Comp. Comp. Comp. Comp.

A3 B3 C3 - A3 B3 C3 - A3 B3 C3 -

Module B2 Comp.
C1-1
N1 EoL1 Comp.
C1-1
N1

Comp. Comp.
C1-2
N2 C1-2
N2
Comp. Comp. Comp. Comp.
B1-1 C1-3
N3 B1-1 C1-3
N3

Comp. Comp. Comp. Comp.

Comp. N4 Comp. N4
B1-2 C1-4 B1-2 C1-4
A1 A1

Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5
Comp.
A2
Comp.
B2-1
Comp.
C2-1
N5 N4
Comp. Comp.
Comp.
C2-2
N6 Comp.
C2-2
N6
B2-2 B2-2
Comp. Comp.
C2-3
N7 C2-3
N7

Comp. Comp.
C2-4
N8 C2-4
N8

Comp. Comp. Comp. Comp. Comp. Comp.

A3 B3 C3 - A3 B3 C3 -

Figure 4.15 Applying reduction strategies to a system structure network

122 4 Concept for the Optimization of Eco-effectiveness of Product Systems

4.4 Data Management of the Input Data

The target of the data management is to provide the relevant input data for the
optimization algorithm. It is required to transfer the product system’s structure
network into a clearly defined data structure where each node and arc of the net-
work is described and weighted according its financial and environmental impact.
Therefore, this subchapter describes the steps two and three of the framework (see
figure 4.1). Furthermore, the individual selections and additional input parameters
selected from the user via the user interface (see figure 4.2) need to be integrated
into the input data. Finally, an interface for entering the optimization results from
the algorithm has to be defined to analyze the results.
Every available relation between the nodes of the structure network has to
be documented to define the network structure for the optimization algorithm.
Therefore, the network structure of the product system is transferred into a data
set where the relevant information for every node and path are entered. The
required information includes “from node”, “to node”, “financial costs”, “GWP”
and “upper boundary” to define the network. The upper boundary information
defines the capacity of each arc and therefore the availability of each module for
a product fleet.
The given example for the data structure in Table 4.1 is based on the product
unit network of figure 4.8 with the premise that thousand product units shall be
build. A maximum capacity of a thousand is required for a module to be included
into every product unit of the product fleet. An upper boundary of the availability
below this value limits the installation rate of this specific module. Every node
category requires a sum of the upper boundaries of at least a thousand to enable
a network flow of thousand product units from the beginning to the end to ensure
that a total of thousand product units can be produced. A lower sum of the upper
boundaries for a node category results in a limited flow through the network as
not enough modules are available for each product.
Consequently, the network structure and architecture hierarchy levels of the
modules and components are clearly defined. The financial and environmental
costs are fixed for each path. The length of this data table is defined by the
number of node connections as every arrow has to be defined.
Furthermore, the user can define a value for the requested minimum size of
product variants for the calculated results. For a fleet of thousand product units,
the extreme scenario for a solution are a thousand individual product configura-
tions. This implies that several measures or modules are potentially applied only
once. A minimum threshold can be defined that forces the algorithm to choose
solutions where a product configuration appears at least as often as this limit
4.4 Data Management of the Input Data 123

Table 4.1 Data structure table for an exemplary network structure

From node To node Financial costs GWP Upper boundary
A B1 e13 120 kg CO2 eq. 1000
A B2 e12 130 kg CO2 eq. 1000
B1 C1 e11 1,500 kg CO2 eq. 330
B2 C1 e23 220 kg CO2 eq. 330
B1 C2 e22 230 kg CO2 eq. 270
B2 C2 e21 2,500 kg CO2 eq. 270
B1 C3 e33 320 kg CO2 eq. 500
B2 C3 e32 330 kg CO2 eq. 500
C1 D1 e31 3,500 kg CO2 eq. 250
C2 D1 e43 420 kg CO2 eq. 250
C3 D1 e42 430 kg CO2 eq. 250
C1 D2 e41 420 kg CO2 eq. 750
C2 D2 e52 530 kg CO2 eq. 750
C3 D2 e51 5,500 kgCO2 eq. 750

value is required to reduce the effort of producing modules for only one single
utilization. For a fleet of thousand product units and a limit value of 20, a maxi-
mum of 50 different product configurations is allowed, to fulfill the constraint of
having at least 20 units of each configuration. This leads to a simplification of
the transferability to a real-world problem, as the investment costs for modules
with a single implementation otherwise are too high. This threshold value is also
defined by the user and is provided to the algorithm via the input table.
The data management table also provides a prepared area for the results of
the optimization. The algorithm and calculation software writes the results of the
calculated eco-effective pathway through the network into the data table where
it is interpreted by the data management system. The entered pathway through
the network is transferred into the ideal set of module and measure combination,
and the total and average environmental and financial impacts per product system
unit and fleet are expressed.
124 4 Concept for the Optimization of Eco-effectiveness of Product Systems

4.5 Adaption of Shortest Path Algorithms

to the Problem Statement

This subchapter describes step four of the framework that has been introduced
in figure 4.1. At first, graph-theory based algorithms to identify shortest paths
through networks are selected to optimize the eco-effectiveness of a product
system. Such an algorithm needs to be adapted to this purpose to meet all require-
ments of the introduced problem statement of optimizing the product system
configuration to meet the goals of eco-effectiveness. The algorithm takes its data
input from the introduced data structure table for the relations between the avail-
able nodes and the further information that is selected by the user’s preferences.
The algorithm design is based on the principle of the minimum cost flow problem,
considering the given arc weightings, the defined network flow, and the capac-
ity limitations per arc. Programming the algorithm is divided into two separate
aspects: model programming and data provision.
Figure 4.16 shows a simple flow chart that describes the general logic of the
optimization model. As the required input data is provided from the input data
storage that is connected to the user interface, the main model is divided into
three segments, namely: (1) initialization of the model, (2) definition of the main
function and (3) definition of the boundary constraints.
During the initialization of the model, the general layout of the optimization
algorithm is defined. In this part of the algorithm code, the structure of the model
and the role of the individual parameters and values are set. This is done by
defining the number of nodes that built the network, the design of the arc structure
that connects the nodes and the key variables that are relevant for the model, e.g.
the limitations for each measure or the objective values.
In the second segment, the main or objective functions are set up. Here it is
defined which values serve as fixed target values (total costs or total emissions)
and which parameter is the value that is minimized.
In the third segment, the constraints or boundary conditions are defined that
make sure that the optimization model behaves in the way the user is expecting.
Among constraints that keep the optimization model within its own boundaries
(e.g. ingoing flow into each node equals outgoing flow), conditions are also
defined that keep the algorithm within the limitations of each measure. The details
of each segment are described in the following paragraphs.

Segment I: Model programming

Programming the optimization model starts with the definition of the required vari-
ables and figures. The total amount of nodes within the network is defined as an
4.5 Adaption of Shortest Path Algorithms to the Problem Statement 125

input data
e.g. number of nodes,
measures, emissions,
costs, limits, …

input data
storage
start

segment I: Initialization of model

e.g. number of nodes, definition of ranges per measure, definition of limitation, definition of
arc design and network flow, definition of target values

segment II: Definition of main function for target optimization

e.g. minimalization of total costs or minimalization of total emissions

segment III: Consideration of constraints and boundary conditions

e.g. compliance to limitation of measures, consideration of upper limit for total emissions or
total costs

end

output data
storage
Output data
e.g. ideal flow, total
emissions, total costs,
…

Figure 4.16 Flow chart of the general logic of the optimization model
126 4 Concept for the Optimization of Eco-effectiveness of Product Systems

integer while the order of the nodes is defined as the range of available nodes. The
options for the optimization objective are defined as float variables and contain the
maximum budget, the maximum emission level along the entire life cycle and the
maximum emission level of e.g. the use stage of the product system. A financial
penalty is introduced that is activated when the emission limit is exceeded to make
sure that the algorithm will select the solution that stays within the given boundaries
or to model existing legislative regulations. For the integration of a minimum size
of configured product types to ensure a minimum amount of product units that are
built in the same way, a “BigM” variable is defined. This variable is complemented
by a variable for the minimum configuration size and upper limit for the number of
possible variations.
These definitions need to be defined within the program code. Figure 4.17 shows
an example of how the definitions of the key variables are translated into the syntax
of the optimization model.

// number and range of all nodes

int NumNodes = ...;
range Nodes = 1..NumNodes;

// limit of target values

float LimitMoney = ...;
float LimitEmissions = ...;
float LimitEmissionsUseStage = ...;

// value of emission penalty

float EmissionPenalty = ...;

// variables for MinConfigurations + types of car bodies

float BigM = ...;
float MinConfiguration = ...;
float MaxCarBodyTypes = ...;

Figure 4.17 Definition of key variables and arc design

In the same way, variables for each node have to be defined. As shown in
figure 4.14 (left), each node, even though it represents the same data, has multi-
ple subnodes to be able to track each selected path individually. For each defined
node, a range is defined that covers these nodes of a module or measure. For the use
case of optimizing multiple product units, availability limits of the defined nodes
become relevant. For each measure, and thus the respective node range, a limit value
is defined that defines the upper boundary for the corresponding arcs.
4.5 Adaption of Shortest Path Algorithms to the Problem Statement 127

The translation of the input data into a bundled information of each arc is done by
the definition of the general arc design. At first, the supply and demand (“SupDem”)
of each node regarding the network flow needs to be defined. This SupDem defines
the individual flow for each node of the network. For the first node, this SupDem
defines the number of product units flowing into the network. The SupDem of the
last node has the same value with a negative sign. All nodes between the first and
last node have a SupDem of zero. This ensures that the amount of product units
that enter a node are the same amount that leaves the node. The tuple design for the
arcs of the network contains the information “fromnode”, “tonode”, “costmoney”,
“costemissions”, and “upper boundary”. This information is directly taken from
the data structure table. The flow through every node has to meet the limits of its
capacity. Those definitions need to be defined within the program code as well.
Figure 4.18 shows an example of how these parts of the program are translated into
the syntax of the optimization model.

// definition of supply and demand

int SupDem[Nodes] = ...;

// definition of arc architecture and source of arcs

tuple arc {
key int fromnode;
key int tonode;
float costmoney;
float costemissions;
int ub; }
{arc} Arcs = ...;
{arc} ArcsUseStage = ...;

// flow in arc needs to stay below upper boundary

dvar int Flow[a in Arcs] in 0 .. a.ub;

// index variable defined for each arc as binary variable

dvar boolean index[a in Arcs];

// calculation of SupDem
subject to {
forall (i in Nodes)
ctNodeFlow:
sum (<i,j,cm,ce,ub> in Arcs) Flow[<i,j,cm,ce,ub>] - sum
(<j,i,cm,ce,ub> in Arcs) Flow[<j,i,cm,ce,ub>] == SupDem[i];

Figure 4.18 Definition of arc design

128 4 Concept for the Optimization of Eco-effectiveness of Product Systems

Segment II: Definition of main function for target optimization

After the variables and the arc design are defined, the equations for the optimization
objectives have to be defined. The objective of the optimization is either the mini-
mization of total costs, the minimization of total emissions or possibly the additional
limitations of the emissions of a single life cycle stage, e.g. the use stage. The total
costs and emissions are calculated by multiplying the individual costs and emissions
of the respective arc with the flow that is assigned to this arc. The optimization objec-
tives are either the minimization of the total emission or the total costs that consist of
the costs of the network flow and the penalty for exceeded emissions. Again, these
definitions need to be defined within the program code as well. Figure 4.19 shows
an example of how these optimization objectives are translated into the syntax of
the optimization model.

// total cost of ownership defined as the sum of all arcs

dexpr float TotalFlowCostMoney = sum (a in Arcs) a.costmoney *
Flow[a];

// total emissions defined as the sum of all arcs

dexpr float TotalFlowCostEmissions = sum (a in Arcs)
a.costemissions * Flow[a];

// total emissions in use stage defined as the sum of all arcs

dexpr float TotalFlowCostEmissionsUseStage = sum (a in
ArcsUseStage) a.costemissions * Flow[a];

// objective function of minimization of target values

Minimize totalFlowCostMoney + EmissionPenalty *
exceededemissions;

Figure 4.19 Definition of optimization objectives

Segment III: Consideration of constraints and boundary conditions

Additional equations for the constraints are required to make sure that the iden-
tified solutions are in line with the defined limitations and boundary conditions.
These constraints define the solution space for the algorithm and set the borders for
the solution identification. Regarding the optimization objectives, constraints are
required to limit the possible results to the threshold values defined by the user. The
first three constraints in figure 4.20 force the algorithm to find solutions that stay
below the user-defined limit values for emissions or costs. The algorithm either has
to choose arcs with lower costs and/or emissions or assign a lower flow through the
more expensive nodes to meet its limitations to stay below the limit values.
4.5 Adaption of Shortest Path Algorithms to the Problem Statement 129

The constraint of providing configuration solutions that appear at least in a min-

imum (and user-defined) amount, is based on two different equations. The first
equation ensures that the flow of a node is at least as large as the minimum con-
figuration variable demands and that the binary index variable of that node has to
be activated. The second equation defines that for enabling a flow within a node,
the index variable has to be activated. This is done by the “BigM” variable which
simply represents a large value. The index variable needs to be activated to make
sure the equation of the flow is being smaller than the BigM variable.
The variable for module limitation controls the limitation of an arc’s availability
within the network flow. This variable determines that the sum of all flows within
the range of the same module is not allowed to be above the user-set capacity limit
of this specific module. Those definitions also need to be defined within the program
code. Figure 4.20 shows an example of how these constraints are translated into the
syntax of the optimization model.

// total emissions need to be below defined limit

ctTotalFlowCostEmissions:
sum (a in Arcs) a.costemissions * Flow[a] <=
LimitEmissions+exceededemissions;

// total costs need to be below defined limit

ctTotalFlowCostMoney:
sum (a in Arcs) a.costmoney * Flow[a] <= LimitMoney;

// use stage emissions need tob e below defined limit

ctTotalFlowCostEmissionsUseStage:
sum (a in ArcsUseStage) a.costemissions * Flow[a] <=
LimitEmissionsUseStage;

// Flow on arcs only when index = 1

forall(a in Arcs)
ctArcUndershoot:
Flow[a] >= MinConfiguration*index[a];

// index variable defines flow

forall(a in Arcs)
ctFlowIndex:
Flow[a] <= BigM*index[a];

// exemplary constraint for module limitations

ctModuleLimitation:
sum (a in Arcs: a.tonode in ModuleRange) (Flow[a]) <=
ModuleLimit;

Figure 4.20 Definition of constraints

130 4 Concept for the Optimization of Eco-effectiveness of Product Systems

Input data provision into the algorithm

The previous section explains the basic formulation of the optimization algorithm
with the required definitions of variables, arc designs, optimization objectives and
constraints. This generic optimization algorithm needs to be filled with fitting data
sets to receive specific results. The data input is done by the data provision interface
of the optimization algorithm and the solving software. Essential input information
is the total number of nodes that the network consists of, the input data regarding
the arc design and the user-selected optimization objective limits. The input of data
is done with an interface that is linked to e.g. Microsoft Excel. By establishing
a connection to a specific excel document, the algorithm takes the required data
from defined areas of this excel document. The input data for each arc is directly
derived from the data structure table (see Table 4.1). Among the arc information,
the SupDem for each node is also provided as well as the relevant user-selected
values for the targeted financial or environmental limitations. In addition to those
variables, the dimension of the penalty for exceeded emissions is also defined.
The provision of the necessary input data has to be included within the program
code. Figure 4.21 shows an example of how the links to the data sources are translated
into the syntax of the optimization model.

// number of nodes in network

NumNodes = value;

// connection to Microsoft Excel

SheetConnection sheet("data.xlsx");

// input from arc information

Arcs from SheetRead(sheet,"Input");
ArcsUseStage from SheetRead(sheet,"Input");
// input of SupDem
SupDem from SheetRead(sheet,"Input");

// input of objective limitations

LimitEmissions from SheetRead(sheet,"Input");
LimitEmissionsUseStage from SheetRead(sheet,"Input");
LimitMoney from SheetRead(sheet,"Input");

// input of emission penalty costs

EmissionPenalty from SheetRead(sheet,"Input");

Figure 4.21 Exemplary provision of input data (I)

Further required input data that can directly be taken out of the data structure
table are the upper boundaries for each arc and module. The limit of each module is
4.5 Adaption of Shortest Path Algorithms to the Problem Statement 131

represented by the capacity of each arc. For each of them, there is an individual limit
and therefore an individual line in the excel sheet. The input value for controlling the
minimum number of produced configurations per type is also directly transferred
to the algorithm. The same applies for the BigM variable which is required to steer
the index variable for each arc to decide whether the flow has to be activated or not.
The provision of the necessary input data has to be included within the program
code. Figure 4.22 shows an example of how the links to the data sources are translated
into the syntax of the optimization model.

// Input of limits
ModuleLimit from SheetRead(sheet,"Input");
Module_II_Limit from SheetRead(sheet,"Input");
Module_III_Limit from SheetRead(sheet,"Input");

// input of size for MinConfiguration

MinConfiguration from SheetRead(sheet,"Input");

// input of BigM
BigM from SheetRead(sheet,"Input");

Figure 4.22 Exemplary provision of input data (II)

Considering the input data, the algorithm can be applied to identify the ideal set
of product system configurations within the defined solution space of the input data
and the constraints.

Output data export

The calculated results are then directly transferred to the excel sheet. The transferred
results contain the value of exceeded emissions, the total amount of emissions along
the entire life cycle and the use stage and the total costs that arise with the selected
product configuration(s). Those values cover all configured product systems and
serve as an overall result. However, the results do not provide details with regard
to the configuration of every single product system of a product fleet. Therefore,
the algorithm also exports the index variable (arc activated or not) and the flow for
each arc. This information is used to trace the flow through the network and identify
the selected configurations and their number. How the links to the data output are
included into the syntax of the program code is shown in figure 4.23 in an exemplary
way.
132 4 Concept for the Optimization of Eco-effectiveness of Product Systems

// result export to Microsoft Excel sheet

exceededemissions to SheetWrite(sheet,"Results!B4");
TotalFlowCostEmissions to SheetWrite(sheet,"Results!B2");
TotalFlowCostEmissionsUseStage to SheetWrite(sheet,"Results!B3");
TotalFlowCostMoney to SheetWrite(sheet,"Results!F2");
index to SheetWrite(sheet,"Results!B8:B386");
Flow to SheetWrite(sheet,"Results!C8:C386");

Figure 4.23 Export of results

4.6 Visualization and Interpretation of Results

As stated above, the results of the algorithm are transferred to a list with the same
structure as the data structure table. In this way, the flow for every individual
arc is identified and the flow through the whole network is reconstructed. This
subchapter describes the fifth step of the framework which has been introduced
in figure 4.1.
In addition to the network flow, the total results for emissions and costs are
available for the product system or the entire product system fleet. When the
ideal flow is identified, it is entered into the product system’s structure network to
visualize the selected path (single product optimization) or the selected modules
(multiple product optimization). Figure 4.24 shows an example of how the ideal
path that is selected by the algorithm through the network for the initial product
system of figure 4.9 is visualized in the product system’s structure network. The
top network in figure 4.24 (network 1) shows the visualization of the optimiza-
tion result for a single product system. The selected components and modules are
highlighted in dark grey to indicate the shortest path regarding either the costs or
the emission level depending on the optimization objective. The bottom network
(network 2) shows the visualization of the results when multiple product systems
flow through the network. Considering multiple product systems, most often sev-
eral different paths exists while each path carries a different share of the total
flow through the network. By incorporating different shades of grey, the different
utilization ratios of each path can be shown. A darker path of connected modules
indicates a higher utilization of this path compared to the modules with a lighter
color.
This kind of result visualization is helpful to understand the results of the
algorithm and to derive decisions regarding the selection and production of indi-
vidual components and modules. Based on this analysis, it can easily be decided
which modules need to be produced in higher or lower volumes. Also, it becomes
4.6 Visualization and Interpretation of Results 133

visible which modules and components do not participate in the ideal path and,
consequently, do not need to be produced or purchased at all. On the other hand,
this visualization does not provide information regarding the number of individ-
ual product configurations and the exact number of each configuration type. The
financial or environmental improvement of the optimized product portfolio is also
not directly visible.

modular product system structure network: Single unit

modular vehicle
1
module A1: combuson engine module B1: conv. car body
Comp. Comp. Comp. Comp. Comp. Comp.
CA1 CB1 CC1 CA1 CB1 CC1
Comp. Comp. Comp. Comp. Comp. Comp.
CA2 CB2 CC2 CA2 CB2 CC2
producon stage P1 end of life stage EoL1
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L1 E1 M1 CA3 CB3 CC3 CA3 CB3 CC3 L1 E1 M1
Loc. Electr. Mach. Loc. Electr. Mach.
L2 E2 M2 L2 E2 M2
module A2: electric powertrain module B2: lightweight car body
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L3 E3 M3 CA1 CB1 CC1 CA1 CB1 CC1 L3 E3 M3

Comp. Comp. Comp. Comp. Comp. Comp.

CA2 CB2 CC2 CA2 CB2 CC2
Comp. Comp. Comp. Comp. Comp. Comp.
CA3 CB3 CC3 CA3 CB3 CC3

modular product system structure network: mulple units

modular vehicle
2
module A1: combuson engine module B1: conv. car body
Comp. Comp. Comp. Comp. Comp. Comp.
CA1 CB1 CC1 CA1 CB1 CC1
Comp. Comp. Comp. Comp. Comp. Comp.
CA2 CB2 CC2 CA2 CB2 CC2
producon stage P1 end of life stage EoL1
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L1 E1 M1 CA3 CB3 CC3 CA3 CB3 CC3 L1 E1 M1
Loc. Electr. Mach. Loc. Electr. Mach.
L2 E2 M2 L2 E2 M2
module A2: electric powertrain module B2: lightweight car body
Loc. Electr. Mach. Comp. Comp. Comp. Comp. Comp. Comp. Loc. Electr. Mach.
L3 E3 M3 CA1 CB1 CC1 CA1 CB1 CC1 L3 E3 M3

Comp. Comp. Comp. Comp. Comp. Comp.

CA2 CB2 CC2 CA2 CB2 CC2
Comp. Comp. Comp. Comp. Comp. Comp.
CA3 CB3 CC3 CA3 CB3 CC3

Figure 4.24 Visualization of results in the network

An alternative display of results shows the individual configurations regarding

their emission level and total costs over the life cycle in a scatter plot. Figure 4.25
shows the general structure of this kind of visualization. On the x-axis, the param-
eter of total costs is represented, the y-axis shows the corresponding value for
134 4 Concept for the Optimization of Eco-effectiveness of Product Systems

total emissions. This creates a field of possible costs and emissions relations
where each configuration fits in.

100

90

80 Area I Area II
70
Total emissions [%]

60

50

40

30
Area III Area IV
20

10

0
0 10 20 30 40 50 60 70 80 90 100

Total costs [%]

Figure 4.25 Scatter plot for visualization of results

Product configurations that fall into “Area I” are configurations with high
emissions, but low costs. Results in “Area II” have an equivalent level of high
emissions, but have higher total costs making these products less desirable. Con-
figurations in “Area IV” have the same level of costs as product systems in “Area
II”, but lower emissions. The most desirable combination of costs and emissions
is found in “Area III” where low emission levels meet low total costs.
Figure 4.26 and figure 4.27 show an imaginary example in order to explain
how this type of scatter plot visualization is used. The example is based on a fic-
tive product system that is produced in over 1,000 different configurations. The
following figures show more than 1,000 different dots within the graph that each
represent a unique emission to cost profile. The distribution of the financial and
environmental impacts can be seen. However, information about which compo-
nents each point consists of is missing. The dashed line in both figures represents
the efficient frontier that marks the ideal solution for the given emission to cost
ratio.
In figure 4.26, a result is shown for the fictitious product system with the
target of optimizing one single product unit. The target is the reduction from the
starting point of 70% emission level (striped star) to the target level of 35% (full
4.6 Visualization and Interpretation of Results 135

star). The result (full star) shows the most economical solution for this limit value
as all other options are either above the emission limit (above the target value
line) or more expensive than this solution (on the right-hand side of the solution
star).

100

90

80

70
Total emissions [%]

60

50

40

30 Target value
20

10

0
0 10 20 30 40 50 60 70 80 90 100

Total costs [%]

Figure 4.26 Scatter plot visualization of results for single product optimization

When e.g. 100 units of the product systems are considered for the optimization
(serving as a fleet of the product system), the solution is visualized as shown in
figure 4.27. The composition of the ideal product fleet is seen in the highlighted
shares of the configuration alternatives. The share of each configuration within
the total product fleet is indicated by the size of the circle around each alternative.
In this example, the ideal fleet consists of eight different product configurations
with different volume shares within the fleet. When those eight configurations are
combined in the indicate volume shares, the resulting fleet average over all units
leads to a more cost-efficient solution than the single unit optimization.
This visualization helps to understand how the algorithm selects the configu-
rations that are part of the ideal fleet. All chosen configurations lie along a fictive
line that marks the pareto optimum. All configurations that are more expensive or
cause more emissions than the configurations on this line will most likely never
be part of the solution.
136 4 Concept for the Optimization of Eco-effectiveness of Product Systems

The calculated results of the algorithm and their interpretation depend on the
form of visualization and the scope of interest. Different types of solution visual-
ization can be created as all relevant solution information is available in the data
table where the flow for every arc is defined. If the focus lies on highlighting the
selected modules and the ideal path, the visualization in figure 4.26 is suitable.
If the focus lies on the fleet structure and the emission and cost profiles are more
relevant, the visualization in figure 4.27 should be preferred.

100

90

80

70
Total emissions [%]

60

50

40

30 Target value
20

10

0
0 10 20 30 40 50 60 70 80 90 100

Total costs [%]

Figure 4.27 Scatter plot visualization of results for multiple product optimization

In general, the results show which product variant(s) are part of the solution.
An optimized fleet composition, consisting of many different alternatives, often
leads to better results than producing only the solution of the single product opti-
mization on a larger scale. Those kinds of visualization help to better understand
the solution and support the decision finding during product development.
Prototypical Implementation
and Application of the Methodology 5

The methodology that is developed in the previous chapter is described in a

generic way to explain the principles of data calculation, data management and
the required steps to identify the ideal product system configuration. The indi-
vidual steps within the optimization framework as shown in section 4.2 are
performed with many different tools and software. Before the introduced method-
ology is applied to a case study, the individual steps and processes are assigned
to suited software tools in order to be able to perform the optimization process to
identify the ideal product system configuration in terms of eco-effective products.
Afterwards, an exemplary application cycle of how to use the implemented tools
is described.

5.1 Prototypical Implementation of the Optimization

Approach

In the following section, the implementation of several software tools for each
step of the framework from section 4.2 is shown. This prototypical implementa-
tion provides one possible way of operationalizing the methodology. Figure 5.1
shows a summary of the software tools that are used for the implementation of
the individual steps of the introduced methodology.

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 137

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_5
 138

Microso Excel
5

Internal TCO Tool,

MatLab Simulink, GaBi

Microso Excel

IBM ILOG CPLEX

Microso Excel,
Microso PowerPoint
Prototypical Implementation and Application of the Methodology

Figure 5.1 Summary of implemented software tools to cover all required steps of the approach
5.1 Prototypical Implementation of the Optimization Approach 139

User interface of the optimization tool

The user interface provides a platform for the user to enter and adjust relevant
information that are needed for the optimization of the product system configura-
tion. Here, the optimization parameters and boundary conditions like optimization
objective, fleet size, measure or module limitations etc. are defined and provided to
an excel sheet in Microsoft Excel. The user can either manually enter the relevant
input data or select them from predefined drop-down menus. Figure 5.2 shows a
part of the exemplary user interface in Microsoft Excel.

Data calculation
The calculation of the input data in terms of environmental impact can be done by
using a specific software for the modelling of product systems with integrated data
bases for the environmental impacts of the processes and materials, e.g. GaBi by
Sphera or Umberto LCA + by ifu hamburg (Broch 2017). The calculation of the
combination-specific data regarding e.g. the individual energy consumption due to
the interdependencies is done by using the Matlab / Simulink software by Math-
Works. The calculation of the total cost of ownership based on life cycle costing
can be done with various calculation softwares or also with Microsoft Excel. For
the following use case, a proprietary calculation software that was programmed for
this purpose is used to calculate and provide the cost input data.

Product system and data management

Structuring of the product system in terms of product architecture of the compo-
nents and modules, transfer to the network structure and information regarding the
connective arcs etc. is done again with Microsoft Excel. Following the design of the
data structure table, the information is entered into the calculation sheet including
the results of the emission and cost impacts and the upper boundaries. The excel
sheet is then used to transfer the data into the software that processes the algorithm.

Processing the input data

For the optimization of the product system configuration based on the input data, a
tool is required that processes the given input data and performs the optimization
algorithm that is defined by the equations of section 4.5. The program ILOG CPLEX
Optimization Studio by IBM is the software that is used in this study. It processes
all input data and transfers the results back to the excel sheet.
140 5 Prototypical Implementation and Application of the Methodology

Figure 5.2 Exemplary user interface for data provision to the optimization algorithm

Visualization
The interpretation and visualization of the results is performed using Microsoft
Excel and Microsoft PowerPoint. The data structure table and the exported results
serve as a basis for the diagrams.
5.2 Exemplary Application Cycle of the Optimization Approach 141

5.2 Exemplary Application Cycle of the Optimization

Approach

The user of the introduced optimization tool has to follow several steps and enter
relevant data and information into this new tool to obtain the desired results. Only
if the user goes through a full application cycle, the desired decision support is
provided. In the context of the automotive industry, the exemplary application
cycle is defined by the optimization of a vehicle fleet. This section provides
a step-by-step guide through the implementation to explain the process from
the user’s perspective following the four segments of the user interface (see
Figure 4.2):

I: Definition of the optimization objective

1. Defining the optimization target, threshold values and penalties

II: Configuration of the product system

2. Defining of the product system and alternative choices along the life cycle
3. Modelling the product system network structure
4. Defining of the vehicle fleet

III: Input data and definition of the financial time perspective

5. Calculating the required input data of LCA and LCC

6. Assigning LCA and LCC results to the product system structure
7. Defining the optimization scenario

IV: Calculation and visualization of the results

8. Setting up the optimization software and data transfer for input data and result
output
9. Interpreting the optimization results

In the user interface, the user needs to decide between the financial and the
environmental indicator as the optimization objective. The chosen indicator is
then minimized and the other dimension defines the threshold value that must
not be exceeded. The appropriate threshold value has to be defined by the user.
Depending on the selected category, the threshold value is either a financial limit
142 5 Prototypical Implementation and Application of the Methodology

(e.g. maximum life cycle cost value per product unit) or an environmental limit
(e.g. emitting a maximum amount of GWP over life cycle per product unit). If, in
the next step, the optimization of a product fleet is selected, the defined threshold
value defines the average limit for the product fleet. Individual units of the fleet
may be above this value. Concerning the definition of the target value, additional
sub-targets for e.g. the use stage can be determined. This stage is additionally
regulated for vehicles by specific legislation as shown in section 2.3. In this
way, the optimization algorithm is able find solutions that comply with use stage
targets and life cycle targets at the same time.
The second step in the application cycle is the definition of the product system
that is to be analyzed. The user has to define which vehicle and which alternatives
or options are the basis for the algorithm to scan through. Therefore, detailed
knowledge regarding the development and production of the vehicle is required.
It has to be coupled with the knowledge about functional alternatives for the
most relevant parts of the vehicle. All relevant decisions with regard to materials,
production processes etc. have to be analyzed and complemented with alternatives
that provide the same function to the vehicle, but with differences regarding GWP
or cost impacts. This can be done with e.g. a bill of materials for the vehicle and
an analysis of the life cycle of a vehicle in terms of its production steps, use stage
and end of life treatment. The outcome of this step has to be a clearly defined
vehicle (e.g. a specific model of a specific OEM, like a Volkswagen Golf VII)
with a clearly defined production process and clearly defined alternatives that
potentially help to improve the environmental performance of the vehicle along
the life cycle.
The third step in the application cycle is the transfer of the vehicle’s architec-
ture, the related production steps, and all alternatives into a network structure in
order to define the directed graph the algorithm can search through. It is important
to define the correct linkages between the individual node categories and alterna-
tives to make sure that only technically feasible paths through the network occur.
The definition of the network structure is done by filling the data structure table
with the respective links regarding the connection of nodes (“from node” and “to
node”). The filled data structure table transforms the vehicle, its life cycle, and
alternative options into a network structure. The data structure table is filled in
the Microsoft Excel sheet that serves as data input for the optimization software.
The fourth step in the application cycle is the definition of the fleet size. The
user needs to define the fleet size by indicating how many vehicles the fleet
comprises. It is also possible to set the fleet size to one in order to optimize a
single vehicle unit.
5.2 Exemplary Application Cycle of the Optimization Approach 143

In the fifth step in the application cycle, the required input data of all nodes
are calculated (by LCA and LCC) that are defined to represent the vehicle’s life
cycle in the product system network. This step requires fundamental knowledge
of the vehicle’s architecture and functionality and also of the calculation of LCA
and LCC results. This step also requires research regarding the financial and
environmental impacts of alternative measures that are not yet applied to the given
vehicle or the entire automotive industry. Every single component and module as
well as life cycle choices need to be assessed individually to be part of the
modular LCA and LCC calculation of the entire vehicle by the algorithm.
The sixth step in the application cycle requires the assignment of the calculated
LCA and LCC results to the network. Each result has to be assigned to the node
it describes. The LCC and LCA results become the weights on the arcs of the
weighted and directed network. This step is done in the Microsoft Excel sheet
where the data structure table is stored.
When the entire vehicle and the alternative measures are modelled and char-
acterized with input data, the user has to define surrounding parameters in order
to define the optimization scenario that the user is interested in. Those parame-
ters are e.g. the expected financial development of crucial influencing parameters
such as material prices and prices for energy and fuels. Here, the user chooses
between different possible scenarios covering several possible developments of
each parameter that has a big influence on the life cycle costs. The user has to
enter current prices as a basis for calculating the trends of the prices as these
volatile parameters can change significantly regarding the location or year the
user wants to simulate. Additionally, the user can optionally define limitations
regarding the availability of each measure that is defined within the structure net-
work. For example: for the alternative powertrain of an electrified vehicle, the
user has the option to limit the availability of green electricity for the use stage
or recycled material for the battery production to simulate different market sce-
narios. By setting different limitations, the user can examine different strategies.
By deleting all limitations, the optimization algorithm finds a fleet that represents
the ideal solution without any constraints to availability. This helps to identify
long-term strategies for the automotive industry. The limitation is defined in the
user interface.
Once the user interface is filled out, the selected optimization software has
to be set up and coupled with the Microsoft Excel sheet that contains the data
structure table for the data transfer. Based on the input data, the optimization
algorithm identifies the ideal flow through the network, i.e. the ideal configuration
and composition of the vehicle fleet.
144 5 Prototypical Implementation and Application of the Methodology

The output of the optimization algorithm is exported in the user interface in

the excel sheet to enable a numerical and a visual presentation of the results. In
the numerical results, the selected alternatives and modules for each function of
the vehicle fleet are listed as well as the (average) emissions and costs per prod-
uct unit. In the visual result presentation, the network of the product system is
depicted, showing the selected alternative for each category of the network. Addi-
tionally, for product fleets, a scatter plot is shown that displays the distribution of
the individual product units regarding their individual environmental and finan-
cial impacts. The average product unit is highlighted as the selected optimization
target. Due to this information, the user can understand which configuration of a
vehicle and its share within the fleet is required to result in the aspired optimized
vehicle fleet.
Application of the Optimization
Approach to a Case Study 6
of the Automotive Industry

In this chapter, the developed optimization methodology is applied to a case

study of the automotive industry. The automotive industry is chosen as it has a
big impact on the environment (as shown in Section 2.3). This industry is driven
by multiple changes in terms of choice of powertrain, digitalization, and innova-
tive mobility solutions such as car sharing, ride pooling or autonomous driving.
This wide-ranging transformation of the industry enables it to integrate new tech-
nologies or approaches into the product design to improve the eco-effectiveness
of a vehicle or fleet. Economically, the automotive industry is also an extremely
large industry. It is responsible for approx. 7% of the European Union’s GDP and
with approx. 14 million people working in the automotive industry, it accounts
for ca. 6% of all jobs within the EU (ACEA 2020). Cost reductions on a large
scale in such an economically important industry can therefore lead to enormous
savings.
A vehicle is a complex product that offers many different module alternatives
and additional measures to possibly alter the specifications of a vehicle along its
life cycle stages. The alternatives result from the fact that a vehicle consists of
thousands of different components and is used for a long and energy-intensive use
stage. This makes it an interesting product system to apply the developed opti-
mization methodology to. The fleet-based recommendation, which is one outcome
of the developed and introduced optimization approach, is also highly relevant
for the automotive industry due to the existing regulations for fleets. The detailed
measures and calculations of this methodology are applied to the vehicle “Golf
VII” by the automotive brand Volkswagen as this vehicle has been produced in a
large number and with various types of powertrains making it an ideal basis for
different scenarios to be compared with each other.

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 145

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_6
146 6 Application of the Optimization Approach to a Case Study …

6.1 Life Cycle Perspectives of a Vehicle

In this subchapter, the typical life cycle of a vehicle is described. Environmental

impacts are analyzed by means of an LCA and the total costs of ownership by
means of an LCC. The obtained information serves as the foundation for the
selection of alternative modules as input data.

6.1.1 Product Life Cycle of a Vehicle

The life cycle of a vehicle is described by the life cycle concept in Section 2.2,
consisting of a production stage, a use stage, and an end of life stage. Prior to
the production stage comes a development phase that can last two to six years.
(Keoleian et al. 1997) Figure 6.1 shows how the generic life cycle of a product
system is transferred to the life cycle of a vehicle.

Figure 6.1 Life cycle of a vehicle

During the production stage, the required raw materials are extracted, pro-
cessed, and transported to the factories where (intermediate) products or parts are
manufactured and finally assembled as a whole vehicle. As nearly three quarters
of all materials are metals (iron, steel, light metals, and nonferrous metals), it is
the most used material in vehicle production. The second biggest group of mate-
rials are polymers with an approx. share of 22% with regard to vehicle weight.
Concerning the raw material extraction, the main processes are energy-intensive
activities such as mining activities, e.g. drilling, refining, smelting, or molding.
6.1 Life Cycle Perspectives of a Vehicle 147

(Keoleian et al. 1997) The average composition of a vehicle with a combustion

engine regarding material use is depicted in Figure 6.2.

2%

4%

22%

3%
63%
6%

Iron & steel Light metals Non ferrous metals Polymers Operating materials Other

Figure 6.2 Share of materials within a vehicle (Volkswagen AG 2017, S. 65)

During the production stage, the components and parts are produced by suppli-
ers or the OEM themselves and assembled to create the final vehicle. Depending
on the vehicle type and the level of detail, a vehicle consists of 10,000 to 40,000
individual parts. (Paasch & Mueller 2013) Typical activities during the produc-
tion and the assembly of those parts are e.g. metal forming, extrusion or forging
as well as joining, welding, fastening, or painting. These processes typically hap-
pen at different locations and require the transportation of parts. (Keoleian et al.
1997).
The use stage is dominated by the operation of the vehicle. In Germany,
the average use stage of a vehicle lasts 13 years. During this time, approx.
170,000 km are covered (Umweltbundesamt 2016). In more detail, the average
lifetime and mileage of a vehicle depends on the type of powertrain and the type
of vehicle itself. Weymar and Finkbeiner (2016) have shown that vehicles pow-
ered by diesel typically have a higher mileage at the end of their use stage than
gasoline driven engines. Also, the covered mileage rises with a higher segment
class of the vehicle. While diesel-powered vehicles cover approx. 180,000 to
148 6 Application of the Optimization Approach to a Case Study …

340,000 miles in 14 years, gasoline-powered vehicles cover 150,000 to 210,000

miles in 18 years. Finally, Weymar and Finkbeiner (2016) propose an aver-
age mileage of 200,000 km per vehicle as the average life cycle performance
for calculations and comparisons of different types of vehicles and powertrains
(Weymar & Finkbeiner 2016). Consequently, this mileage is considered in this
study.
Concerning the use stage, the most important aspect is the provision and
consumption of fuel or other energy carriers for the operation of the vehi-
cle. Additionally, other materials such as operational fuels or replacement parts
are consumed. Furthermore, surrounding services like repairs, maintenance or
insurances are offered and often required during the use stage. (Keoleian et al.
1997).
At the end of the life cycle, the vehicle is being dismantled for the reuse of
certain parts, the recycling of materials or the disposal of waste and scrap mate-
rials. Figure 6.3 shows this end of life treatment process for vehicles. The take
back of a vehicle begins with the depolution where hazardous materials or liquids
(e.g. batteries, airbags, fuels etc.) are removed. Then, different components (e.g.
tires, glass, etc.) are disassembled that are not suited for shredding of the vehicle
as they cannot be separated afterwards. Those parts are then either prepared for
reuse or recycling. The remains of the vehicle are shredded and sorted—using
sieving or magnetic separation processes—into three different categories: ferrous
metals, non-ferrous metals, and other materials. While the metals are mostly recy-
cled, the other materials (mainly polymers) are usually burnt or sent to landfill.
(Umweltbundesamt 2019b).

6.1.2 Environmental Assessment of a Vehicle’s Life Cycle

Due to the complex production stage, the long and energy-intensive use stage
and the challenging end of life stage, a vehicle is the cause for many envi-
ronmental impacts along its life cycle. This leads to environmental damages on
different scales. The functional unit is defined as the transport of passengers over
200,000 km within the driving profile of a defined test cycle (e.g. NEDC or
WLTP) to assess and compare these environmental influences. Possible environ-
mental impact categories that are relevant for the vehicle’s life cycle are e.g. the
GWP, the eutrophication potential, the ozone depletion potential, and the acidi-
fication potential. In the following, the developed methodology is applied to the
category of GWP for the vehicle “Golf VII” by Volkswagen based on the results
6.1 Life Cycle Perspectives of a Vehicle 149

End-of-life
Depolution Dismantling Shredding
vehicle

Registration, Oils, batteries,

unlocking and cooling materials and Parts, materials and Iron, (nonferrous)
emptying airbags recyclable Materials metals and residues

Storage and Recycling Combustion

disposal and sale and landfill

Figure 6.3 End of life treatment of vehicles

of an LCA following the principles of the ISO 14040 standard (as described in
Section 2.2).
The environmental profile of a Golf VII regarding its greenhouse gas emis-
sions over the entire life cycle highly depends on the choice of the powertrain
and the further configurations of vehicle aspects, e.g. equipment for infotainment
etc. The Golf VII has been available with five different powertrain types:

• Combustion engine powered by gasoline

• Combustion engine powered by diesel
• Combustion engine powered by compressed natural gas (CNG)
• Plug-In hybrid powertrain of a gasoline engine and a battery electric engine
• Battery electric vehicle (BEV)

Assuming an identical optional vehicle equipment for all five vehicles (regarding
choices for entertainment etc.), every powertrain choice results in an individual
GWP profile over life cycle.
The conventional gasoline vehicle reaches a total amount of approx. 34.4 t of
GHG emissions over its life cycle and therefore has the highest impact on the
GWP of all powertrain types. Those 34.4 t of GHG emissions are divided into
approx. 5.2 t of CO2 eq. occurring during material extraction, part production and
vehicle assembly, 5.4 t of CO2 eq. coming from fuel provision and approx. 23.8
t of CO2 eq. occur during the use stage due to the combustion of gasoline. The
150 6 Application of the Optimization Approach to a Case Study …

diesel driven Golf VII results in a total of 28 t of GHG emissions. The production
stage causes 5.8 t of CO2 eq. , fuel provision is responsible for 2.2 t of CO2 eq. and
the use stage for 20 t of CO2 eq . The CNG powered engine respectively results
in 5.8 t of CO2 eq. , 5.4 t of CO2 eq. and 18.8 t of CO2 eq. , resulting in 30 t of
GHG over its life cycle. The emission profile of a BEV Golf VII differs from
the profile of a conventional combustion engine. The 23.8 t of CO2 eq. over its
life cycle are separated into 11.4 t of CO2 eq. for vehicle production and 12.4 t of
CO2 eq. for electricity provision (assuming the European average energy mix from
2017). (Volkswagen AG 2019b) For a plug-in hybrid electric vehicle (PHEV), the
total emissions sum up to 27.4 t of CO2 Eq. 8.2 t of CO2 eq. are caused during the
production stage, 11 t of CO2 eq. during fuel provision and electricity production
(again assuming the European average energy mix from 2017) and 8.2 t of CO2 eq.
during the use stage when gasoline is combusted. The different GHG profiles for
the different powertrain types are illustrated in Figure 6.4, showing that the choice
of powertrain highly influences the environmental impact of a vehicle.
Diesel Gasoline

∑ = 34,4t

∑ = 28,0t
CNG

∑ = 30,0t
electric range: 45 km, EU electricity mix of 2017
PHEV

∑ = 27,4t
range: 250 km, EU electricity mix of 2017
BEV

∑ = 23,8t

0 5 10 15 20 25 30 35
producon well to tank tank to wheel recycling [t CO2-eq. / life cycle]

Figure 6.4 Greenhouse gas emissions of different types of powertrains (based on Volkswa-
gen 2019b)

Analyzing the relative distribution of the total emissions per vehicle for the
individual life cycle stages, the difference between combustion engines and elec-
trified powertrains becomes even more visible. For a combustion engine, approx.
15% of the emissions occur during the production stage, 15% to 20% during fuel
provision and with 60% to 70% most of the emissions are caused during the use
stage. Emission reduction techniques for combustion engines therefore mainly
focus on the reduction of fuel consumption in order to reduce emissions. For a
6.1 Life Cycle Perspectives of a Vehicle 151

BEV, nearly 50% of the emissions are caused during the vehicle production, the
other half is caused during electricity generation, assuming the average electricity
mix of Europe. The high share of the production stage for BEVs is caused by a
reduced impact of the use stage and an increased impact during the production
stage compared to combustion-based vehicles (see Figure 6.4). If an electric vehi-
cle is powered by electricity coming from renewable sources, the level of GHG
emissions is reduced to approx. 12 t of CO2 eq. making the production stage the
main source of emissions (with over 95%). The share of emissions for each life
cycle stage is shown in Figure 6.5.
The environmental hotspot for combustion engines is the use stage. It can be
influenced by either reducing the energy demand or changing the type of fuel
(e.g. from gasoline to diesel or CNG). For electrified powertrains, the hotspots
are electricity provision for the use stage and the production of the battery system.
Reduction potentials can be realized by applying renewable energy sources to the
use stage and e.g. the battery cell production.
Diesel Gasoline

∑ = 34,4t

∑ = 28,0t
CNG

∑ = 30,0t
electric range: 45 km, EU electricity mix of 2017
PHEV

∑ = 27,4t
range: 250 km, EU electricity mix of 2017
BEV

∑ = 23,8t

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
producon well to tank tank to wheel recycling [t CO2-eq. / life cycle]

Figure 6.5 Distribution of emission sources along the life cycle stages (based on Volkswa-
gen 2019b)

6.1.3 Total Cost of Ownership along a Vehicle’s Life Cycle

Parallel to the ecological life cycle assessment, the five different powertrain alter-
natives of the Golf VII are assessed regarding their total cost of ownership. The
scope is set as introduced in Section 2.2 and contains the costs of the produc-
tion stage (customer price), use stage costs and end of life costs. Following the
152 6 Application of the Optimization Approach to a Case Study …

customer’s perspective on the TCO, the purchasing costs of the vehicle sum-
marize the costs of the production stage (combining material costs, production
costs, development costs and distribution costs) and the OEM’s profit. The use
stage costs represent the costs of the owner of the vehicle who operates it for
200,000 km. Included are the costs for the operating materials expenses for
maintenance and service as well as taxes and insurance.
The costs for acquisition and end of life treatment can be quantified follow-
ing the OEM’s information and the given assumptions. The costs for the use
stage, however, are more complicated to calculate as the use stage can last up
to 20 years, in which the costs for variable expenses such as taxes, insurance,
services or fuels etc. can vary a lot. Due to this reason, the costs of the use
stage include a correcting factor for inflation adjustment and the optimization
tool offers a choice of different cost development scenarios.

Costs of production stage

The costs for the production stage from the customer’s point of view can be taken
from the general information of the vehicles OEM. The selling price of each vehicle
mostly depends on the choice of powertrain, the choice of special equipment and
further additional configurations and vehicle extras. Comparing the basic configu-
ration, the gasoline powered Golf VII offers the lowest production costs, followed
by diesel, CNG, BEV and last PHEV. However, the basic configurations of each
vehicle are not comparable regarding their different trim levels. Since e.g. the e-
Golf contains more features and special equipment in its basic configuration than
the basic gasoline driven Golf, the functional unit of these two vehicles differs if the
trim level is included into the scope of financial comparison. If the level of extras
is equalized and adjusted by configuring the vehicle, e.g. the gasoline driven vehi-
cle’s extra equipment configured to the level of the basic e-Golf, the financial gap
between the powertrains shrinks drastically. Table 6.1 shows the purchasing costs
from the customer’s perspective for the five different types of powertrains in terms
of basic configuration and purchasing price if the trim level is adjusted to a common
level of equipment.
In the further analysis of the vehicles and the optimization of their eco-
effectiveness, the adjusted prices are selected to be the respective production costs
for each powertrain type to eliminate the influence of different trim levels. Fur-
thermore, it is assumed that the customer pays the whole purchasing price in total
without any financing model or discount.
6.1 Life Cycle Perspectives of a Vehicle 153

Table 6.1 Purchasing costs of different powertrains (Volkswagen AG 2019a)

Type of powertrain Basic price Adjusted price based on configuration
Gasoline e19,300 e26,600
Diesel e23,875 e29,200
CNG e26,575 e30,515
PHEV e36,900 e31,575
BEV e35,900 e35,900

Costs of use stage

The costs of the use stage of the different types of powertrains depend on the
expenses for fuel, service and maintenance, taxes, and insurance. Costs for fuel
or electric energy depend on the specific price of the energy carrier and the con-
sumption of this energy carrier over lifetime. For the gasoline driven vehicle, a
consumption of 4.9 l per 100 km is assumed (Volkswagen AG 2019a), resulting in
a fuel consumption of 9,800 L of gasoline for 200,000 km over lifetime. Assum-
ing an average gasoline price of e1.5 per liter over lifetime, the expenses for fuel
result in e14,700 over lifetime for the gasoline powered Golf VII. Analogous to
this calculation scheme, the expenses for fuel for the diesel driven vehicle result in
e10,500 (4.2 l per 100 km and e1.25 assumed average diesel price) and for the
CNG powered vehicle in e9,144 (3.6 kg of CNG with an assumed average price
of e1.27 per kg of CNG over life time) (Volkswagen AG 2019a). For the BEV, an
energy consumption of 14 kWh per 100 km is expected. Considering losses during
the charging process of approx. 11%, the energy demand rises to 15.73 kWh per
100 km. Calculating with an assumed average price of e0.3 per kWh (Bundesnetza-
gentur & Bundeskartellamt 2021, p. 288), the expenses for energy result in e9,438
over lifetime. For the PHEV, a fuel consumption of 3.3 l of gasoline and 7 kWh
of electric energy (7.86 kWh including losses) are assumed for 100 km resulting
in fuel expenses of e14,619 over lifetime (Volkswagen AG 2019a). The costs for
service and maintenance, taxes and insurances are based on assumptions and data
from the ADAC (German Automobile Club) (ADAC 2019a). Table 6.2 shows the
individual costs and the cumulated costs of the use stage for the different types of
powertrains. These values do not necessarily represent the true costs over the entire
use stage, but act as the use stage representation of this case application and the
following optimization.
154 6 Application of the Optimization Approach to a Case Study …

Table 6.2 Assumed costs of use stage for different powertrains

Type of Fuel Service & Taxes Insurance Total costs of
powertrain maintenance use stage
Gasoline e14,700 e8,063 e1,260 e2,370 e26,393
Diesel e10,500 e8,063 e3,300 e2,900 e24,763
CNG e9,144 e9,753 e960 e2,640 e22,497
PHEV e14,619 e8,399 e420 e2,370 e25,808
BEV e9,438 e7,315 -e e2,370 e19,123

Costs of end of life stage

Due to the expectation that after 200,000 km the vehicle reaches its end of life stage,
it is further assumed that the vehicle does not have a residual value anymore and is
shredded. The end of life treatment, as shown in Figure 6.3, is assumed to be equally
cost intensive for all powertrain types, with each costing e500.

Comparison of TCOs for different powertrains

Figure 6.6 shows the TCO for the different types of powertrains in comparison.
The results show that while the BEV has the highest purchasing costs, the TCO are
still lower than for the PHEV as the use stage costs are lower than for any other
powertrain. The most expensive vehicle over its life cycle is the PHEV as it combines
the second highest costs during the production stage with the highest costs of the use
stage. Given the introduced parameters and assumptions, the lowest TCOs in this
study are reached with either a gasoline engine or a CNG driven vehicle. However,
with growing costs for exhaust gas treatment for combustion based vehicles and
expected sinking costs for battery cells, the cost structure will most like change in
future. This change will then most likely result in financial advantages for electrified
vehicles.
As stated before, the BEV has the highest share of production costs, but also
the lowest share in the use stage costs. The analysis of cost distribution roughly
corresponds to the distribution of the GHG emissions of the different powertrains
which is shown in Figure 6.5. The TCO results for the five given powertrain types
serve as the basis for further analysis and optimization of the vehicle configurations.
Figure 6.7 shows the distribution of the TCO along the different life cycle stages
for each type of powertrain.
6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 155
Diesel Gasoline

∑ = 53,5k €

∑ = 54,4k €
CNG

∑ = 53,5k €
electric range: 45 km
PHEV

∑ = 57,8k €
range: 250 km
BEV

∑ = 55,5k €

0€ 7,500 € 15,000 € 22,500 € 30,000 € 37,500 € 45,000 € 52,500 € 60,000 €

producon use end of life [€ / life cycle]

Figure 6.6 Total costs of ownership from customer’s perspective

Diesel Gasoline

∑ = 53,5k €

∑ = 54,4k €
CNG

∑ = 53,5k €
Electric Range: 45 km
PHEV

∑ = 57,8k €
Range: 250 km
BEV

∑ = 55,5k €

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
producon use end of Life [€ / life cycle]

Figure 6.7 Distribution of the TCO along the different life cycle stages

6.2 Selection of Measures for the Reduction

of Greenhouse Gas Emissions

In Sections 6.1.2 and 6.1.3, the LCAs and LCCs for five different types of
powertrains in their initial configuration have been presented. The five vehicle
alternatives represent the baseline scenarios for all further vehicle configurations
that are created by the optimization algorithm in the further progress. Additional
measures or module alternatives are required that help to reduce the level of GHG
156 6 Application of the Optimization Approach to a Case Study …

emissions, and in the best case also decrease the TCO. If that is not possible, the
extra measures should at least increase the TCO of a vehicle as little as possible.
As previously shown, many GHG emissions occur and vary over different
life cycle stages due to different processes. Often, the emission profile directly
depends on the choice of powertrain. For each powertrain and life cycle stage,
different approaches and measures exist that help to reduce the emissions. Fur-
thermore, there are also measures that affect all powertrain types in the same
way. In this chapter, possible measures and their alternatives are identified that
potentially help to reduce GHG emissions during the life cycle of a vehicle. As
changes in the product configuration can lead to interdependencies, the effects
of alternatives on the different types of powertrains or other modules are also
analyzed. Furthermore, the effect on LCA and LCC is provided for the selected
measures and modules.

6.2.1 Selection of Measures and Module Alternatives

The reduction of GHG emissions for vehicles has usually been done by reducing
the fuel consumption during the use stage since this life cycle stage (from well
to wheel) is the source for approx. 85% of the vehicle’s emission if a combustion
engine is used (O’Reilly et al. 2016). While for vehicles with powertrains fueled
by petrol-based energy carriers the reduction of energy demand during the use
stage is very effective, alternative powertrains show different emission hotspots
that require different measures.
As shown in Section 6.1.2, GHG emissions occur in all life cycle stages, which
also enables reductions in all stages. Alternative materials or material sources as
well as secondary materials coming from recycling offer opportunities to lower
the GHG emissions in the production stage. Further measures are innovative pro-
duction processes, a higher energy efficiency during vehicle production or use of
renewable energy carriers to power the factories. Emission reductions during the
use stage are either achieved, as stated above, by reducing the energy demand or
by using energy carriers with a reduced emission profile (Harvey 2018; O’Reilly
et al. 2016).
Three different options are available to reduce the energy demand of a vehicle
during the use stage: (1) increasing the efficiency of the engine, (2) reducing the
energy losses during the transmission of the energy from engine to wheels and
(3) reducing the physical driving resistances. Those driving resistances are shown
in equation 6.1, where Pe represents the effective engine power, PDTL the power
losses of the drive train, PESL the power losses of engine slip and PDRL the power
6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 157

losses of the driving resistances (Schuetz 2016, p. 11 ff.).

Pe = PDT L + PE S L + PD R L (6.1)

The power losses due to the forces of the driving resistances (F DRL ) in turn
comprise the forces of the rolling resistance (F RR ), the acceleration resistance
(F AR ), the aerodynamic resistance (F AER ), and the gradient resistance (F GR ) (see
equation 6.2) (Schuetz 2016, p. 11).

FD R L = FR R + FA R + FAE R + FG R (6.2)

The calculation of the individual resistance forces is shown in equations 6.3 to

6.6. Equation 6.3 shows the calculation of the force of rolling resistance (F RR ).
The variable cR represents the coefficient of the rolling resistance, mv represents
the mass of the vehicle and g the force of gravity. (Schuetz 2016, p. 11)

FR R = c R ∗ m v ∗ g (6.3)

Equation 6.4 shows the calculation of the force of acceleration resistance (F AR ).

The variable i represents the coefficient of the mass inertia, mv represents the
mass of the vehicle and a the acceleration of the vehicle. (Schuetz 2016, p. 11)

FA R = i ∗ m v ∗ a (6.4)

Equation 6.5 shows the calculation of the force of aerodynamic resistance (F AER ).
The variable cw represents the coefficient of the flow resistance, A represents the
surface area of the vehicle, ρ the air density and vr2el the relative velocity of the
vehicle. (Schuetz 2016, p. 11)

ρ ∗ vr2el
FAE R = cw ∗ A ∗ (6.5)
2

Equation 6.6 shows the calculation of the force of gradient resistance (F GR ). The
variable sin α represents the gradient of the track, mv represents the mass of the
vehicle and g the force of gravity. (Schuetz 2016, p. 11)
158 6 Application of the Optimization Approach to a Case Study …

FG R = sin α ∗ m v ∗ g (6.6)

Analyzing the formulas for the calculation of the driving resistances, the relevant
parameters for the energy demand during the use stage are the mass of the vehicle,
the front area of the vehicle and the coefficients for flow and roll resistance of
the vehicle. Summarizing the different sources of direct and indirect emission
sources along the automotive life cycle, the following technological areas offer
measures and alternatives to reduce the emission profile of a vehicle (see also
Figure 6.8):

• Raw materials: alternative material sources, utilization of alternative materials

• Production: innovative processes, higher energy efficiency, renewable energy
carriers
• Use: Alternative fuels, alternative powertrains, green energy carriers, weight
reduction, reduced roll resistance, reduced flow resistance, reduced vehicle
front area, higher efficiency of engine, higher efficiency of energy transmission
within the vehicle
• End of life: provision of secondary materials

producon use phase recycling

producon: landing gear: closed loop economy:

• energy efficiency • roll resistance reduced res • Li-Ion-baery recycling
• green energy • opmised recycling
• innovave processes • FVK recycling
body frame & interieur:
• aluminum recycling
• renewable raw materials
• efficient air condioning
• aerodynamics opmisaon
• light weight construcon: Steel
/ Aluminium / CFK / …

powertrain & fuels:

• BEV, PHEV, CNG, diesel,
gasoline, …
• green electricity, eFuels, …
• opmized engine &
transmission efficiency

Figure 6.8 Possible emission reduction approaches

6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 159

Those technological areas show the fields in which measures and alternatives
can possibly be applied to reduce emissions. Still, they do not show how exactly
the targeted measure can be implemented. Within every technological area and
for each step of the life cycle, detailed measures are required to reach the emis-
sion benefits. For the approach of e.g. vehicle mass reduction, there are many
ways to achieve a reduced vehicle weight. Possible measures range from the use
of different materials (aluminum, carbon fiber etc.) to different geometries that
require less material for the same function. Within each implementation branch
of measures, e.g. light weight construction using aluminum instead of steel, many
subversions are possible. The options range from replacing 1% up to nearly 100%
of steel with aluminum.
The developed methodology is applied to a use case of selected measures
that cover a broad variety of options and alternatives for emission reduction.
These measures include alternative powertrains as well as raw material supply,
vehicle production, physical aspects regarding driving resistances and transport
and logistics. In detail, the selected measures are:

• Raw materials: five different material options for the car body: (1) 100% steel
and (2-5) steel and aluminum in four varying amounts for different weight
reductions
• Production: green electricity for battery cell production, foil instead of
varnish for colored vehicle parts, transport by ship instead of train
• Use: powertrains based on diesel, gasoline, CNG, plug-in hybrid (PHEV) and
battery electricity (BEV), regenerative CNG alternative (eGas), regenerative
electricity for BEV and PHEV, optimized aerodynamic body kit, optimized
tires with reduced roll resistance
• End of life: secondary materials for cathode production of batteries for electric
vehicles instead of primary materials

These measures are set up as a product system network that is used as a solution
space for the optimization algorithm. The selected exemplary measures are suffi-
cient to cover all typical influences on the vehicle’s energy consumption during
the use stage and the emission sources during the production stage. The design
of this network is shown in Figure 6.9. It shows that there are 20 different pow-
ertrain and fuel configurations that can be combined with 20 different vehicle
bases and four different options for production and logistics. In total, this net-
work of measures and optional modules offers 1,600 different paths from the
“begin” node to the “end” node for one single vehicle. For a fleet of 100,000
 160

20x powertrains & fuels 20x vehicle base 4x producon &

6

Regular logiscs
CNG
CNG- Steel
powertrain carbody
eGas

Steel &
Gasoline-
Gasoline alminum
powertrain
carbody
Regular Regular Regular Transport
res aerody. varnish ship
Aluminum
Diesel-
Diesel & steel
powertrain
End

carbody

Begin
Opm. Opm. Foil for Transport
res aerody. varnish train
Alminum
PHEV- Regular Regular Use: EU
& steel
powertrain materials producon mix
carbody

BEV- Recycled Green Use: Green Alminum

powertrain materials producon mix carbody

Figure 6.9 Module network for case study application

Application of the Optimization Approach to a Case Study …
6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 161

vehicles, this network results in 1,600 different ways of configuring each vehi-
cle for the fleet leading to 1,600100,000 options for the fleet composition, given a
100% availability of every module.
Possible measures for the reduction of a vehicle’s GHG emission that are not
considered in this study are a change in the mobility behavior of the vehicle
owner and the geometry of the vehicle to a point where the functional unit is
changed (e.g. changing the car body to a smaller vehicle with a lower passenger
capacity).

6.2.2 Measure Analysis Regarding LCA, LCC and Availability

The LCA and LCC results for the five basic powertrain types are shown in
Section 6.1.2 and 6.1.3. The results already include the regular modules of the
vehicle’s “standard” configuration. For the results of e.g. the e-Golf, the costs
and LCA results regarding the following measures are already included: regular
material option for the battery cells, regular energy mix for the battery production
and use stage, steel car body, regular tires and aerodynamic kit, regular varnish,
and transport via ship. For the modular optimization of the vehicle configuration,
the LCC and LCA results of the vehicle have to be split up to the individual
modules of the regular vehicle configuration. All costs and emissions that cannot
be directly associated to a measure or module are included in the costs of the
powertrain which serves as the “vehicle basis”. Costs and emissions that fall into
this category come e.g. from the window glass, the seats or other components that
are not individually included in the exemplary network of this case study. The
financial expenses for maintenance, service, taxes, insurance etc. are included in
the powertrain aspect as they are not represented in the exemplary network at
any other stage. The optional modules also need to be assessed regarding LCA
and LCC to serve as input variables for the algorithm. Finally, the given modules
need to be assessed in terms of their individual availability for the incorporation
into a fleet of multiple vehicles (in this example a fleet of 100,000 VW Golf VII).
Table 6.3 shows the costs, emissions and availability of the basic vehicle con-
figuration and the respective energy carriers and fuels. The costs and emissions
of the basic vehicles are taken from Sects. 6.1.2 and 6.1.3 and have an exemplary
calculated offset for the additional measures that follow in the module network.
The costs and emissions cover all aspects that are not directly covered within the
other nodes of the network. This is done to create a network that covers all costs,
even though only a limited number of measures and parts are considered. How-
ever, these data do not directly relate to the real module costs of a Volkswagen
162 6 Application of the Optimization Approach to a Case Study …

Golf VII, but are used to simplify the modeling of the network. The availabil-
ity of each powertrain and fuel is based on estimations for illustration of the
functionality of the algorithm. For all combustion-based engines, 100% availabil-
ity is assumed meaning that all 100,000 vehicles of the VW Golf VII fleet can
be built with such a powertrain. An exception is the CNG based engine which
has a reduced availability due to a limited gas station availability. The availabil-
ity for electrified powertrains is reduced to 60% (PHEV) and 40% (BEV) to
increase the complexity for the fleet composition and to show the flexibility of
the optimization approach.

Table 6.3 Summary of

Measure or Financial GWP Availability
input data for basic vehicles
module costs [%]
and energy carriers
Gasoline e35,975* 1.95 t CO2 100%**
powertrain eq.*
Diesel e41,075* 2.55 t CO2 100%**
powertrain eq.*
CNG e41,531* 2.55 t CO2 50%**
powertrain eq.*
PHEV e40,356* 2.95 t CO2 60%**
powertrain eq.*
BEV e43,237* 2.1 t CO2 eq.* 40%**
powertrain
CNG e9,144* 24.2 t CO2 45%**
eq.*
eGas (CNG) e14,328* 5.2 t CO2 eq.* 15%**
(PHEV)
Diesel e10,500* 22.2 t CO2 100%**
eq.*
Gasoline e14,700* 29.2 t CO2 100%**
e9,900* eq.*
(PHEV) 19.7 t CO2
eq.* (PHEV)
Electricity e9,438* 12.4 t CO2 100%**
EU mix (BEV) eq.*
e4,719* 6.2 t CO2 eq.*
(PHEV) (PHEV)
(continued)
6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 163

Table 6.3 (continued)

Measure or Financial GWP Availability
module costs [%]
Green e12,584* 0.4 t CO2 eq.* 15%**
electricity (BEV) 0.2 t CO2 eq.*
e5,977* (PHEV)
(PHEV)
* = Assumptions based on data from Volkswagen (see
Section 6.1.2 and 6.1.3)
** = based on assumptions for illustration

Table 6.4 shows the input data for cost, emission impact and availability of
the production-related measures and alternatives for the exemplary case study. For
the VW e-Golf with its 35.8 kWh battery capacity and the VW Golf GTE with
a battery capacity of 8.7 kWh the costs for the battery are assumed to be 5,000
e (e-Golf) and 1,000 e (Golf GTE), respectively (ADAC 2019b; ADAC 2019c).
These costs are derived from an expected price of 100 e per kWh and total costs
for a battery system of 10,000 e for a capacity of 50 kWh to 70 kWh (Becker
2019). It is assumed that these costs are evenly spread on the costs for the battery
materials and the costs for the electricity of the battery production. The costs for
the more ecological alternatives (secondary battery materials and green electric-
ity for battery production) are expected to be more expensive than the original
alternatives. The environmental impact of the battery related measures are also
derived from the total vehicle LCA of Section 6.1.2. The difference between the
electric vehicle and the gasoline vehicle is expected to be the impact of the bat-
tery and further electric components. The impacts of the more ecological options
are based on exemplary assumptions for the problem design, as well as the data
of the varnish and foil and of the transportation options. These measures and data
are included to represent further life cycle stages within this case study. The avail-
ability of the regular options is set to 100%. This means that 100% of all possible
electrified vehicles can be equipped with primary materials. The availability of
secondary materials is set to only 10% as returning batteries for recycling are not
yet available in large scales. The same applies to renewable energy for battery
production. The availability of the foil and transportation alternatives is set to
35% and 50% to increase the complexity of the fleet configuration.
Table 6.5 shows the input data for cost, emission impact and availability
of measures and alternatives that influence the production stage due to mate-
rial choices, but also the use stage due to influences on the physical driving
164 6 Application of the Optimization Approach to a Case Study …

Table 6.4 Input data for production-focused alternatives

Measure or module Financial costs GWP Availability [%]
Primary battery e2,500* (BEV) 3 t CO2 eq.* (BEV) 100%**
materials e500* (PHEV) 1 t CO2 eq.* (PHEV)
Secondary battery e2,750* (BEV) 1.5 t CO2 eq.* (BEV) 10%**
materials e700* (PHEV) 0.45 t CO2 eq.* (PHEV)
Battery production EU e2,500* (BEV) 3 t CO2 eq.* (BEV) 100%**
electricity mix e500* (PHEV) 1 t CO2 eq.* (PHEV)
Battery production e2,750* (BEV) t CO2 eq.* (BEV) 30%**
green electricity e700* (PHEV) t CO2 eq.* (PHEV)
Regular varnish e375** 0.5 t CO2 eq.** 100%**
Foil for varnish e500** 0.2 t CO2 eq.** 35%**
Transport by ship e150** 0.5 t CO2 eq.** 100%**
Transport by train e450** 0.2 t CO2 eq.** 50%**
* = Assumptions based on data from Volkswagen (see Section 6.1.2 and 6.1.3) and ADAC
** = based on assumptions for illustration

resistances. The costs, environmental impacts and availability of tires and aero-
dynamic options are based on assumptions for the exemplary problem design as
well as the data for the availability of the car body options. The LCA results
for the environmental impact of the five different car body options assume a
total of exemplary 600 kg steel within the entire car body with assumed asso-
ciated cutting losses of exemplary 40% during the production of the car body.
This requires processing 840 kg of steel. The GWP factor of 2.4 kg of CO2 eq.
per kg of steel (Sphera 2019) results in emissions of approx. 2 tons of CO2 eq.
in the standard configuration for the 100% steel-based car body. Considering an
assumed material substitution rate of 50% for aluminum as steel replacements
and a GWP factor of 9.5 kg of CO2 eq. per kg of aluminum (Sphera 2019), the
further configurations have an environmental impact from 2.2 t of CO2 eq. up
to 2.7 t CO2 eq. in combination with 100 kg of weight reduction. The financial
costs are based on assumptions regarding the problem complexity simulation.
The data regarding the energy consumption that are shown in table 6.3 are
taken from Sects. 6.1.2 and 6.1.3 and represent the initial vehicle configuration.
These values only work if the basic configuration is chosen for every option
in the network that influences the fuel consumption. If an alternative option is
included in the vehicle configuration, these values lose their correctness and new
values that represent the individual vehicle specifications need to be calculated
6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 165

Table 6.5 Input data for use stage-focused alternatives

Measure or module Financial costs GWP Availability [%]
100% steel based car body e1,000** 2 t CO2 eq.* 100%**
steel and aluminum based car e1,250** 2.2 t CO2 eq.* 75%**
body (−25 kg)
steel and aluminum based car e1,750** 2.35 t CO2 eq.* 50%**
body (−50 kg)
steel and aluminum based car e2,250** 2.5 t CO2 eq.* 35%**
body (−75 kg)
steel and aluminum based car e2,750** 2.7 t CO2 eq.* 15%**
body (−100 kg)
Regular tires e1,300** 0.25 t CO2 eq.** 100%**
Optimized tires e2,000** 0.5 t CO2 eq.** 65%**
No aerodynamic kit e0** 0 t CO2 eq.** 100%**
Aerodynamic kit e1,000** 0.25 t CO2 eq.** 55%**
* = based on data from Sphera and assumptions
** = based on assumptions for illustration

and included in the data input for the optimization algorithm. As the given options
offer 20 different configurations that change the driving resistances, the five pow-
ertrains that are available in this case study result in 20 individual consumption
patterns. The effect of these interdependencies is explained in the following. It
provides an example regarding the weight reduction by choosing a different car
body material.
As stated before, the reduction of the vehicles mass is a possible option to
reduce the driving resistance of a vehicle and likewise its energy demand and
emission level in the use stage. In the module network of this case study, five
different weight scenarios are adapted as possible design alternatives. The highest
possible delta of vehicle mass in this case study is the switch from the stan-
dard configuration based on a steel car body (600 kg car body weight) to the
lightweight concept car body weight of 500 kg (400 kg of steel and 100 kg of
aluminum).
While the physical energy demand due to the reduced weight is the same
for all vehicle and powertrain concepts, the saved amount of fuel and, thus,
the related GHG emissions are different depending on the detailed powertrain
mechanisms. As an option of the node category “car body concept” needs to be
combined with an option from the powertrain range, those differences need to be
166 6 Application of the Optimization Approach to a Case Study …

considered. The different potentials regarding the reduced emission level of the
same weight reduction for different powertrain types arises as each powertrain
has an individual level of emission for the provision of energy. A combustion
engine based on diesel or gasoline and an electric vehicle show differences in
terms of fuel consumption due to engine efficiencies and fuel characteristics to
provide 1 kWh of energy to the vehicle’s movement. The reduction of the energy
demand has to be calculated and transferred to a reduction of the chosen fuel
(e.g. gasoline or diesel) to calculate the emission reduction. The resulting value
is called “Fuel Reduction Value” (FRV). (Koffler & Rohde-Brandenburger 2010;
Rohde-Brandenburger 2013).
The driving conditions are defined by the vehicle test cycle. Based on these
conditions, the required energy can be calculated (based on the Eqs. (6.3) to (6.6))
and extrapolated for 100 km. The energy demand, combined with the efficiency
losses of the vehicle engine and power transmission, is converted into a fuel
consumption. For a distance of 100 km, driven with the load profile of the new
European driving cycle and a weight reduction of 100 kg of the vehicle mass, the
driving resistances are reduced in such a way that the energy demand is reduced
by approx. 0.54 kWh. According to Rohde-Brandenburger (2013), this reduced
energy demand is converted to a reduced fuel consumption of 0.12 l for diesel
and to 0.15 l for gasoline. All these results are calculated with a simple reduction
of the vehicle weight, but without any adaptions to the powertrain or other parts
of the vehicle to adapt to the reduced weight. If the engine design is adapted to
the reduced weight, e.g. through a reduced cylinder capacity, the reduction of the
fuel consumption can be further improved. Rohde-Brandenburger (2013) indicates
fuel reduction values of approx. 0.35 l per 100 km for the gasoline engine and
approx. 0.28 l per 100 km for the turbocharged diesel engine. (Koffler & Rohde-
Brandenburger 2010; Rohde-Brandenburger 2013).
Those fuel reduction values combined with the emission factors of the given
fuels (2.33 kg CO2 for one liter of gasoline and 2.63 kg CO2 for one liter of
diesel (Wansart 2012, p. 14)) lead to the reduction of CO2 emissions during
the use stage due to the weight reduction by 100 kg. For the gasoline engine,
the reduction of CO2 emissions results in approx. 0.82 kgC O2
100km and for diesel in
approx. 0.74 kgC O2
100km . However, as the emission factors only include the combustion
of the fuel but not the provision of it, the fuel production from oil refinery to the
transportation to the gas stations has to be included as well. For gasoline, the
emission factor for the provision of one liter of gasoline is approx. 0.51 kg CO2
(Sphera 2019). The emission reduction for a gasoline engine given a vehicle
6.2 Selection of Measures for the Reduction of Greenhouse Gas Emissions 167

weight reduction of 100 kg including the fuel provision increases to approx.

0.99 kgC O2
100km . . (Koffler & Rohde-Brandenburger 2010; Rohde-Brandenburger 2013)
For electric vehicles, these fuel reduction values do not apply directly, as the
recuperation of energy during braking processes allows the vehicle to recover
energy. Therefore, the mass of the vehicle has a different impact on the energy
consumption than for combustion-based vehicles. (Vetter 2017) Consequently,
the degree of emission reduction due to mass reduction depends on the choice of
powertrain. If an electric vehicle is powered with renewable energy, the vehicle
has nearly zero emissions during the use stage and the weight reduction does not
have a significant impact at all on the GWP.
The combination of a measure or option that influences the vehicle’s mass
with a powertrain type shows a typical example of the interdependencies of two
measures and the impact these interdependencies can have on the emission reduc-
tion potential. Giving a certain car body option (e.g. weight reduction of 100 kg)
a fixed GWP reduction value leads, in this case, to wrong results for the opti-
mization, as the correct value depends on the powertrain it is combined with.
This interdependency effect only has an impact during the use stage of the vehi-
cle, the production of the car body option or the (unadapted) powertrain are not
influenced by the interdependencies. This allows the proposed exclusion of the
correction factor for the use stage as shown in Section 4.4. An important aspect
of the total reduction of emissions is the impact of the new car body concept in
its production stage. The reduction of emissions during the use stage of reduced
mass is partly mitigated by a higher impact in the production stage due to the
utilization of materials with a higher GWP per kg (e.g. aluminum for steel).
The principle of interdependencies occurring from different weight reduction
options in combination with different powertrains also applies to other measures
that influence the driving resistance, e.g. aerodynamic optimization or a reduced
rolling resistance. Each combination of measures that influence the driving resis-
tances with a certain type of powertrain results in a unique energy demand and
emission level. This unique situation for every combination has to be provided by
the data input to the optimization algorithm. Influences of those interdependen-
cies to the total emissions of GHG and the TCO are considered during the use
stage due to the energy or fuel consumption and for the costs for energy or fuel
provision. Changes regarding service or maintenance costs, taxes or insurance
etc. based on these interdependencies are not considered in this case study.
168 6 Application of the Optimization Approach to a Case Study …

6.3 Data Input for the Vehicle and Fleet Optimization

Price scenarios
The individual data regarding financial costs, environmental impact, and availability
for implementation into vehicle fleets has to be provided to the optimization algo-
rithm as described in Section 4.4. Also, the user-selected boundary conditions need
to be processed and included in the input data calculation. These user-made choices
influence the prices for fuels and electricity during the use stage or influence the
price of raw materials, affecting the price of the vehicle during the production stage.
For the scenarios of fuel and electricity prices, the user enters the current price of
each energy carrier and defines three different target values that set the price for the
end of 2030. The three different price choices represent three different scenarios
(best case, normal case, worst case) for each energy carrier. By selecting one of the
three scenarios, the fuel price linearly grows or shrinks towards the defined target
value in 2030. This represents a changing fuel price. For the material prices, the
cost level of aluminum and steel are defined. A price development until the end of
2030 can be defined by three different scenarios. If the aluminum price is expected
to rise to e.g. 130% of its starting price by 2030 and the steel price is going to rise
to a level of 125%, then these prices will appropriately be included over the time in
the calculation.
The influence on the TCO of increasing fuel prices over time leads to different
optimization results compared to TCO results with static fuel prices of today’s
price levels. A gasoline driven vehicle with a fuel consumption of 4.9 L per 100 km
consumes a total of 9,800 L of gasoline over a use stage of 200,000 km. Considering a
constant price level of 1.50 el , , the total fuel costs amount to e14,700 (see Table 6.2).
A rising gasoline price up to 2.00 el until December 2030 (starting in September
2018) leads to an average fuel price of 1.75 el if the vehicle’s use stage is also
represented in this time span. The total fuel costs then sum up to e17,150, making
this scenario e2,450 more expensive than the scenario with a static fuel price. The
same logic applies to the material prices. A target value of 125% price growth for
steel and 130% price growth for aluminum until 2030 results in an average fleet
price value of 112.5% for steel and 115% for aluminum over time.

Network adaption
Starting from the network of measures for this use case (see Figure 6.9), the indi-
vidual combination options and configuration alternatives need to be modelled into
a network that allows the algorithm to trace individual combination paths (shown in
Figure 4.15). The network is altered, and the order of measures is rearranged to apply
6.3 Data Input for the Vehicle and Fleet Optimization 169

the network to the required structure. The new logic for the network that is given to
the optimization algorithm is shown in Figure 6.10. A total of 212 individual nodes
are required to represent all combination options in a network in an efficient way.
The first and the last node are the beginning and the end of the network and do not
represent a real measure or option. The nodes two to six represent the five different
car body alternatives and the nodes seven to eleven represent the combination of
each car body option with the choice of regular tires. In return, the nodes 12 to 16
stand for the five combinations of the five different car body designs together with
the optimized tires for a reduced rolling resistance. The nodes seven to 16 represent
the five car body options with both tire options. The five paths from nodes two to
six towards the nodes seven to eleven have the same information regarding costs,
emissions, and availability, as all edges lead to the same measure (in this case regular
tires). The nodes 17 to 26 (ten nodes) represent the combinations of the five car body
designs together with regular tires and the option of no aerodynamic optimization.
The nodes 27 to 36 represent the selection of the optimized aerodynamic body kit.
The nodes 17 to 36 (20 nodes) in total express the 20 possible combinations of car
body, tires, and aerodynamics that are displayed in Figure 6.9. The next node cate-
gory (nodes 37 to 196, in total 160 nodes) serves as the combination of all 20 vehicle
configurations with the different available powertrains. The individual nodes of that
category stand for:

• First node group (nodes 37 to 56): all 20 vehicle options for gasoline powertrain
• Second node group (nodes 57 to 76): all 20 vehicle options for diesel powertrain
• Third node group (nodes 77 to 96): all 20 vehicle options for CNG powertrain
• Fourth node group (nodes 97 to 116): all 20 vehicle options for e-Gas powertrain
• Fifth node group (nodes 117 to 136): all 20 vehicle options for PHEV powertrain
• Sixth node group (nodes 137 to 156): all 20 vehicle options for PHEV (green
electricity) powertrain
• Seventh node group (nodes 157 to 176): all 20 vehicle options for BEV
powertrain
• Eighth node group (nodes 177 to 196): all 20 vehicle options for BEV (green
electricity) powertrain

The nodes 197 and 202 are, just like nodes one and 212, not associated to a real
option but serve as a “performance node” to streamline the network design and
reduce the required number of paths. In node 197, all 40 PHEV powertrain variants
are bundled and connected to the following options for battery material choices
(nodes 198 and 199) and electricity mix for battery production (nodes 200 and 201).
In node 202, all 40 options for the BEV configuration are bundled and connected
170 6 Application of the Optimization Approach to a Case Study …

to the following choices for battery material (nodes 203 and 204) and electricity
for battery production (nodes 205 and 206). Following the material and electricity
options for PHEV and BEV within the network, each powertrain offers 160 indi-
vidual configurations. Node 207 serves as a collector of all 400 available vehicle
configurations and bundles them as an input for the last choices of varnish or foil
(nodes 208 and 209) and transport via ship or train (nodes 210 and 211).

37 “Collector”
… node
56
Measure
7 17 node
57
…
8 18 76

9 … 77
…
10 25 96
2

11 26 97
3 …
116 208 210
Begin: End:
4 207
1 212
117 209 211
5 12 27 …
136 198 200
6 13 28 197
137 199 201
…
14 …
156

15 35
157
…
16 36 176 203 205
202
177 204 206
…
196

Figure 6.10 Network logic for optimization network

This network represents the network design that will be searched by the algorithm
to identify the ideal network flow. This network needs to be given to the optimization
software in form of a data structure table as shown in Table 4.1. The adaption of the
data structure table to this specific case study results in an input table with 379 rows
while each row defines one of the 379 required arcs of this network. 166 of these
arcs are edges that do not connect selected choices to further options, but serve as
connections to the collection nodes or the end node. These 166 nodes do not carry
any information regarding emissions, costs or availability and are therefore marked
with costs of e0, emissions of 0 t CO2 eq. and an availability of 100% to make sure
6.4 Results of Optimization for Different Scenarios 171

that the algorithm does not consider any restrictions for these connecting edges.
Figure 6.11 shows the first 120 lines of this data structure table (further lines in
appendix 1). The column for costs represents the individual costs for each node, the
TCO are the sum of all node values of the final vehicle configuration. The emissions
are displayed in grams of CO2 eq. per driven kilometer, assuming a mileage of
200,000 km. The upper boundary is oriented on a fleet size of 100,000 vehicles.
The optimization algorithm for this use case is based on the general code that is
introduced in Section 4.5.2, adjusted to the specific case study. The generic code is
adapted to an exact use case regarding the input data and requires the exact definition
which cell of the excel sheet contains the relevant parameter for e.g. measure of
availability limitations. Also, the number of constraints needs to be adapted to the
given problem statement by adding an individual constraint for each arc limitation
(upper boundaries).

6.4 Results of Optimization for Different Scenarios

The optimization algorithm is applied to the case study of both the vehicle and
fleet optimization. For the single vehicle scenarios, a VW Golf VII represents
the test object. For the fleet optimization, 100,000 VW Golf VII need to be
configured by the optimization algorithm. As shown before, the selected alter-
native modules that are available to modify the basic vehicle result in a total of
1,600 different vehicle options. These 1,600 options are the searching space for
the single vehicle optimization and the vehicle pool to choose from for the fleet
optimization (respecting the given limitations). Figure 6.12 shows these 1,600
vehicle combinations in a scatter plot with the TCO on the x-axis and the level
of emissions (in gram of CO2 eq. per km) on the y-axis.
The plot shows the results of the financial and environmental evaluation as
described in Sections 6.1.2 and Section 6.1.3 expanded by the results of the
alternatives for each powertrain. Each individual dot represents a unique vehicle
configuration and its emission and cost profile. These data are the starting point
for the optimization of different scenarios. In the following subchapters, the opti-
mization results for a single vehicle and vehicle fleets are analyzed according to
their emission levels complying with the following limit values of:
172 6 Application of the Optimization Approach to a Case Study …

Figure 6.11 First 120 lines of data structure table

 emissions [g CO2 eq. / km]
• Gasoline
6.4 Results of Optimization for Different Scenarios

• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

total cost of ownership [€]

Figure 6.12 Emission and TCO profile of all vehicle configurations

173
174 6 Application of the Optimization Approach to a Case Study …

• 175 g of CO2 eq. per km (over the entire life cycle including all life cycle
stages)
• 165 g of CO2 eq. per km (over the entire life cycle including all life cycle
stages)
• 125 g of CO2 eq. per km (over the entire life cycle including all life cycle
stages)
• 105 g of CO2 eq. per km (over the entire life cycle including all life cycle
stages)

6.4.1 Analysis of the Results for a Single Vehicle

Optimization

The optimization for a single vehicle considering the four different emission
limitations is done by the algorithm. However, given all 1,600 options in the
scatter plot in Figure 6.12, the ideal configuration can also be derived from the
plot. Assuming the given prices, emissions, and availabilities as described in the
previous sections and assuming standard development scenarios for fuels and
materials, the solution space results in the vehicle profiles shown in Figure 6.12.
The most cost-efficient vehicles that are identified for each scenario by the algo-
rithm are shown in table 6.6. This table shows the vehicle configuration, the final
emission level and the TCO value for each scenario.
The results show that the algorithm results in a different solution for every
scenario. For a reduced limit value, the previous solution is not valid anymore.
The results also show that for all four scenarios, the only variation is the choice of
powertrain as a reaction on the reduced emission limit. All choices regarding car
body, tires, aerodynamics, varnish, and transport remain in the basic configuration
while the powertrain and fuel choice is changing for each scenario. For the first
scenario (limit = 175 gCkmO2 ), the algorithm results in a gasoline driven vehicle
with an emission level of 172.2 gCkmO2 and a TCO value of e53,447 over lifetime.
For the second scenario (limit = 165 gCkmO2 ), a CNG driven powertrain is recom-
mended with an emission level of 150.7 gCkmO2 and a TCO value of e53,556. For
the third scenario (limit = 125 gCkmO2 ), the algorithm results in a BEV powertrain
with primary battery materials, a regular energy supply for the battery production
and a regular energy supply for the use stage (119.3 gCkmO2 and e55,497 TCO).
For the fourth scenario (limit = 105 gCkmO2 ), the algorithm results in the same
vehicle, but with a renewable energy supply for the battery production (104.8
gC O2
km and e55,747 TCO).
6.4 Results of Optimization for Different Scenarios 175

Table 6.6 Single vehicle optimization results

limit: limit: limit: limit:
175 g CO2 eq. / 165 g CO2 eq. / 125 g CO2 eq. 105 g CO2 eq. /
km km / km km
Powertrain Gasoline CNG BEV, primary BEV, primary
materials & materials &
regular renew. electricity
electricity for for prod
prod
Fuel Gasoline CNG EU mix EU mix
Car body 100% steel 100% steel 100% steel 100% steel
Tires Regular tires Regular tires Regular tires Regular tires
Aerodynamic No aerodynamics No aerodynamics No No aerodynamics
aerodynamics
Varnish Varnish Varnish Varnish Varnish
Transport Ship Ship Ship Ship
Emissions 172.2 g CO2 eq. 150.7 g CO2 eq. 119.3 g CO2 104.8 g CO2 eq. /
/ km / km eq. / km km
TCO e53,447 e53,556 e55,497 e55,747

Figure 6.13 shows how these four vehicles are visualized in the network of
this case study. The red path (1) describes the result for the first scenario, the
green path (2) the results for the second scenario and the black path (3), (4) the
result for scenario three and four. It becomes visible that the only changes for
the adaption to the different emission values occur at the stage of powertrain and
fuel selection. Afterwards, all vehicles follow the same path through the network.
Figure 6.14 shows the solutions of the different scenarios in the scatter plot of
all vehicle alternatives. For the first scenario it is clearly visible that the algorithm
chooses the most cost-efficient of all vehicle options as all vehicles have an
emission level below the first threshold value. As all 1,600 vehicles fulfill the
emission constraint, the algorithm simply chooses the most cost-efficient of all
versions, viz. the regular gasoline vehicle (red circle A). The green circle (B)
marks the recommended vehicle for the second scenario, viz. a CNG powertrain
is chosen. Even though this vehicle has an emission level which is far below the
required limit, it is, regarding its TCO, still more cost-efficient than a gasoline
vehicle that is closer to the limit value.
 176

20x powertrains & fuels 20x vehicle base 4x producon &

2
6

Regular logiscs
2 CNG
1 2 3 4
CNG- Steel
powertrain carbody
eGas
1 1 Steel &
Gasoline-
Gasoline alminum 1 2 3 4
powertrain 1 2 3 4 1 2 3 4 1 2 3 4
carbody
Regular Regular Regular Transport
res aerody. varnish ship
Aluminum
Diesel-
Diesel & steel
powertrain
End

carbody

Begin
Opm. Opm. Foil for Transport
3 4 3 3 4 res aerody. varnish train
Alminum
PHEV- Regular Regular Use: EU
& steel
powertrain materials producon mix
carbody

3 4 4
BEV- Recycled Green Use: Green Alminum
powertrain materials producon mix carbody

Figure 6.13 Network flow for optimized vehicles

Application of the Optimization Approach to a Case Study …
6.4 Results of Optimization for Different Scenarios 177

The two grey circles (E1 and E2) mark the vehicles for the third and fourth
scenario. Both vehicles are closer to their emission limit value than the CNG vehi-
cle in scenario two and are only distinguished between the selected energy mix for
the battery production. For the vehicle selection of scenario three (E1), it becomes
visible how the algorithm chooses a vehicle that undershoots its emission limit
(here a gap of 5.7 gCkmO2 ) to find the most cost-efficient solution. A vehicle that
exactly matches the limit value, e.g. a PHEV-based vehicle, has a TCO value
that is approx. e2,500 more expensive than the selected BEV (indicated with the
orange X).

6.4.2 Analysis of Results for Vehicle Fleet Optimization

The optimization of a vehicle fleet with 100,000 individual vehicles is analyzed

for the same four scenarios as the single vehicle optimization.

First scenario
For the first scenario (limit = 175 gCkmO2 ), the algorithm finds the same solution for
the vehicle fleet as for the single vehicle optimization. The ideal fleet configuration
results in 100% of the basic gasoline vehicle as shown in Table 6.6. The detailed
vehicle configuration is the same as indicated by the red path in Figure 6.13. In
Figure 6.15, the fleet result is shown for the first scenario. The circle around the
vehicle indicates the share of this configuration within the total fleet. In this case,
all 100,000 vehicles are built in the same configuration resulting in a fleet average
of 172.2 gCkmO2 and a TCO value of e53,447 over lifetime per vehicle. The red bar
to the right of the scatter plot also shows the share of the gasoline powertrain within
the fleet (here: 100%).
For the first scenario, no difference to the single vehicle optimization occurs as
the basic gasoline vehicle fulfills the emission limitation in the most cost-efficient
way and no measure limitations reduce the availability of this vehicle configuration.

Second scenario
For the second scenario (limit = 165 gCkmO2 ), the fleet composition of the first scenario
cannot be used anymore, as this fleet does not meet the emission limit. For the new
limit, the algorithm finds a fleet that exactly meets the limit by reaching an average
emission level of 164.9 gCkmO2 . . This fleet has an average TCO value of e53,482
and consists of two different vehicle configurations. The first vehicle type is the
gasoline vehicle (configuration A), which has a fleet share of 67.1% for the new
limit. The rest is based on the basic CNG powertrain configuration which is the same
 scenario 1
178

A scenario 2

X
scenario 3
6

E1

scenario 4
E2

emissions [g CO2 eq. / km]

• Gasoline
• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

total cost of ownership [€]

single vehicle, 175 g CO2 / km: single vehicle, 165 g CO2 / km: single vehicle, 125 g CO2 / km: single vehicle, 105 g CO2 / km:

A = 172.2 g CO2 | 53,447 € B = 150.7 g CO2 | 53,556 € E1 = 119.3 g CO2 | 55,497 € E2 = 104.8 g CO2 | 55,747 €

Figure 6.14 Scatter plot results for single vehicle optimization

Application of the Optimization Approach to a Case Study …
6.4 Results of Optimization for Different Scenarios 179

fleet composion:
100 %
A fleet average: limit 175 g CO2 eq.
= 172.2 g CO2 eq. | 53,447 €
emissions [g CO2 eq. / km]

• Gasoline
• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

0%
total cost of ownership [€]

Figure 6.15 Vehicle fleet optimization for scenario one

vehicle as the solution for the single vehicle optimization of the second scenario
(see Figure 6.16).

fleet composion:
100 %
A fleet average: limit 165 g CO2 eq.
= 164.9 g CO2 eq.| 53,482 €
B
32.9 %
emissions [g CO2 eq. / km]

67.1 %
• Gasoline
• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

0%
total cost of ownership [€]

Figure 6.16 Vehicle fleet optimization for scenario two

The network flow for this scenario is shown in Figure 6.17.

This fleet design results in an average fleet vehicle which is approx. e74 cheaper
regarding the TCO than the single vehicle result.
180 6 Application of the Optimization Approach to a Case Study …

20x powertrains & fuels 20x vehicle base 4x producon &

Regular logiscs
CNG
CNG- Steel
powertrain carbody
eGas

Steel &
Gasoline-
Gasoline alminum
powertrain
carbody
Regular Regular Regular Transport
res aerody. varnish ship
Begin

End
Aluminum
Diesel-
Diesel & steel
powertrain
carbody
Opm. Opm. Foil for Transport
res aerody. varnish train
Alminum
PHEV- Regular Regular Use: EU
& steel
powertrain materials producon mix
carbody

BEV- Recycled Green Use: Green Alminum

powertrain materials producon mix carbody

Figure 6.17 Network flow for fleet optimization scenario two

Third Scenario
The solution of the fleet optimization for the third scenario is more complex. While
the single vehicle optimization identifies the basic BEV configuration as the solution,
the fleet optimization results in a mix of eleven different vehicle configurations
spread over five different types of powertrains. On average, this fleet meets the limit
of 125 gCkmO2 with an average TCO value of e55,005, being approx. e492 cheaper
than the single vehicle result. The ideal fleet composition consists of 38.8% of the
basic CNG vehicle, 15% of the basic diesel configuration and 6.2% of the basic eGas
configuration. 34% of the fleet are based on a BEV powertrain with the regular EU
electricity mix, 6% of the whole fleet are BEVs powered by renewable energy. The
algorithm chooses a total of eight different BEV configurations to be part of the final
fleet, four of them are based on renewable electricity and four on the regular EU
mix. Each group consists of the same four vehicle configurations which are a mix of
BEVs with regular battery materials produced with regular electric energy, regular
battery materials produced with renewable energy for battery production, recycled
battery materials combined with regular energy for battery production and recycled
battery materials and renewable energy for battery production (see Figure 6.18).
The indicated stars in the diagram on the right of Figure 6.18 indicate that this
powertrain is exhausted, as the algorithm has met the limitation of the availability
for this powertrain. This means that even if the algorithm wanted to shift even more
volume towards vehicles based on e.g. eGas, the limitation forbids this utilization
and forces the algorithm to find an alternative for the fleet composition. The added
“Line 1” is an auxiliary line to link the vehicle configuration B (CNG) and G2 (BEV
with renewable energy). This line marks the ideal ratio of emissions and costs for
this scatter plot and indicates the ideal solution for each emission limit which lies
as close as possible to this line. The algorithm chooses vehicles G1 to G4 to be part
6.4 Results of Optimization for Different Scenarios 181

fleet composion:
100 %
fleet average: limit 125 g CO2 eq.
* 6.2 %

B
= 124.9 g CO2 eq.| 55,005 € * 6.0 %

C
emissions [g CO2 eq. / km]

E
4
E
* 34.0 %

3
E
1
E
2

15.0 %

G
4
H
•
G
Gasoline 3
• CNG G
1
• Diesel G
2
•
•
PHEV
BEV * 38.8 %

• PHEV – Green Power

• CNG – eGas
• BEV – Green Power

0%
total cost of ownership [€]

Figure 6.18 Vehicle fleet optimization for scenario three

of the ideal fleet over any vehicle with a PHEV powertrain as all other vehicles are
further away from this ideal line and therefore relatively more expensive for their
level of emissions than other vehicles.
Figure 6.19 shows the network flow for the measure choices for the third scenario.
The red stars in this figure indicate the exhausted options due to measure availability
limitations.

20x powertrains & fuels 20x vehicle base 4x producon &

38.8 % logiscs
Regular CNG
45 % CNG-
* 100 % Steel
carbody

*
powertrain

6.2 % eGas

Steel &
Gasoline-
Gasoline alminum
powertrain
carbody
100 % 100 % 100 %
100 %
Regular Regular Transport
Regular res
Begin

aerody. varnish ship

End

Aluminum
15 % Diesel-
15 % Diesel & steel
powertrain
carbody
Opm. Opm. Foil for Transport
res aerody. varnish train
Alminum
PHEV- 36 % Regular 28 % Regular 34 % Use: EU
materials producon mix
& steel
powertrain
carbody

40 % BEV-
powertrain
4 % Recycled
materials
12 % Green
producon
6%
Use: Green
mix
* Alminum
carbody

Figure 6.19 Network flow for fleet optimization scenario three

For the automotive industry, not only the total emissions over the entire life cycle
are relevant, but also the emissions that occur only during the use stage (here: well
to wheel perspective). The calculation of an optimized fleet with a limit of 125 gCkmO2
182 6 Application of the Optimization Approach to a Case Study …

over the entire life cycle with an additional emission limit of 95 gCkmO2 during the use
stage (from well to wheel) leads to a different fleet design as now two target values
need to be fulfilled. Figure 6.20 shows the result for this adapted optimization target
for the same life cycle limit with the sub-limit for the use stage.

fleet composion:
100 %
emission limit life cycle: 125 g CO2 eq. 6.8 %
emission limit use phase: 95 g CO2 eq. *
B = 124.9 g CO2 eq.| 55,208 €
* 6.0 %

C
2

*
emissions [g CO2 eq. / km]

34.0 %
E
4

26 %
G
4
H
• Gasoline
• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power 27.2 %
• CNG – eGas
• BEV – Green Power

0%
total cost of ownership [€]

Figure 6.20 Vehicle fleet optimization for scenario three with use stage limits

The algorithm needs to include more diesel vehicles and less CNG vehicles
into the fleet to lower the use stage emissions to stay below the limit value of 95
gC O2
km during the use and fuel provision stage. In fact, this is done as the diesel
vehicle has lower emissions in its tank to wheel profile than the CNG vehicle (see
Figure 6.4). Considering this adapted fleet, the new fleet average meets the stricter
emission limits, showing higher costs with an average TCO value of e55,208 (+
e203 compared to the basic version of the third scenario).

Fourth scenario
The fourth scenario with a limit of 105 gCkmO2 as the average fleet emission for
all 100,000 vehicles is the most complex limit to reach for the algorithm from all
analyzed scenarios. The final solution consists of a total of 116 individual vehicle
types. This fleet results in a TCO value of e57,996 over lifetime and an average
emission level of 108.9 gCkmO2 . . This result means that the algorithm is not able to find
a solution that respects the given vehicle options and limitations while remaining
below the demanded emission limit. Due to the extra high penalty costs for every
exceeded gram of emissions over the limit value (here: e100,000 per exceeded gram
and vehicle), the algorithm still tries to find a fleet composition that is as close to the
6.4 Results of Optimization for Different Scenarios 183

limit as possible in order to avoid the penalty for exceeding the limit. The final fleet
consists of the following powertrains: 4 × CNG with eGas, 32 × PHEV, 16 × PHEV
with renewable electricity, 48 × BEV and 16 × BEV with renewable electricity (see
Figure 6.21).
As visible in the bar chart on the right of Figure 6.21, the algorithm has exceeded
the availability of all used powertrains except for the regular PHEV configuration.
The algorithm uses all possible BEV configurations that are below (or close to) the
limit and close to the auxiliary line to come as close as possible to the emission limit.
When all these options are exceeded, the algorithm selects all available vehicles that
are based on BEVs with renewable electricity or CNG vehicles with eGas as they
are below the emission limit and close to the auxiliary line that marks the ideal
emission in terms of cost ration. When these options are not available anymore, the
algorithm includes the PHEV vehicles with renewable electricity. Even though they
are far away from the auxiliary line, the algorithm still chooses them in order to
stay below the limit value. Exceeding these options, the algorithm fills up the fleet
with all possible left-over BEV and PHEV configurations that are above the limit.
Thus, the algorithm has chosen all vehicle options that help to create a fleet with
the lowest emission profile possible. Yet, it is not able to meet the limit value.

fleet composion:
100 %
fleet average: limit 105 g CO2 eq.
* 6.8 %

= 108.9 g CO2 eq.| 57,996 €

* 6.0 %

D
D5 D D
D D
emissions [g CO2 eq. / km]

D 6D
5 7
6 7 D 8 D 1D D D D D DDD D D 34.0 %
E E E EE EE
4 4 E444E E
E
E EE
5 5E 5444 EE44 E4
4E E 4E 4
E
8 1 2D 3 D D
4 5 6 7
D
2 3 D 4 5D 6D 7 8 8 1 1D 2 D
2 3 D4
D D
3
4
*
3 3 E35E3E E
6 6 6 E
E E E EE 3EE
E 6 EE E
1 1 E111E E
E
7 7 E7111 EE
11 E1 E
1 E EE
E11 E E1 E1
2 2 7 2E EE E F
8 8 872 E 7 E E E 8 5
F F F
2 6 7 1 F F
8 8 88 F 2 3 F
F F 9.0 %
F F4

*
8 5 F F
1
6 7 F 2 3 F
8 4

G G
4 4G G
H5 5
1 HG HG H
• Gasoline 2 3 G4 G
3 3
6 6
• CNG G G
1 1G G
• Diesel 7 G7 G
G 44.2 %
2 2G
• PHEV 8 8

• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

0%
total cost of ownership [€]

Figure 6.21 Vehicle fleet optimization for scenario four

Figure 6.22 shows the network flow for the fourth scenario.
The red stars mark the modules that have met their individual limit of availability.
While Figure 6.21 only shows the vehicle configurations sorted by powertrain, the
network flow in Figure 6.22 also shows the utilization of further measures such
184 6 Application of the Optimization Approach to a Case Study …

20x powertrains & fuels 20x vehicle base 4x producon &

Regular logiscs
CNG
6.8 % CNG- 65 % Steel
carbody

*
powertrain

6.8 % eGas

Steel &
Gasoline-
Gasoline alminum
powertrain carbody
50 %
35 % Regular 45 % Regular 65 % Regular
Transport
res aerody. varnish
Begin

ship

End
Aluminum &
Diesel-
Diesel steel
powertrain

* * * *
carbody
50 %
65 % Opm. 55 % Opm. 35 % Foil for
Transport
res aerody. varnish
train
53.2 %
PHEV-
powertrain
83.2 %
Regular
materials
63.2 %
Regular
producon
78.2 % Use:
EU mix
35 % Alminum &
steel carbody *
40 % BEV-
powertrain * 10 %
Recycled
materials
* 30 % Green
producon * 15 %
Use: Green
mix
* Alminum
carbody

Figure 6.22 Network flow for fleet optimization scenario four

as car body options or tire alternatives. In Figure 6.22, it becomes visible that not
only the powertrain options of CNG with eGas, BEVs with regular and renewable
electricity and PHEVs with renewable electricity are exhausted, but also some of
the additional measures.
In detail, the algorithm uses all available vehicle configurations that contain the
emission reducing measures, e.g. transport via train instead of transport via ship or
foil instead of regular varnish, and also includes all available optimized aerodynamic
options and optimized tire options. Furthermore, the algorithm includes all available
vehicles with a weight-reduced car body (− 75 kg), which all are included in the
regular PHEV vehicles. For the electrified vehicles, all options for recycled battery
materials and use of renewable energy for battery production are used.

6.4.3 Sensitivity Analysis of Results

For the sensitivity analysis of several parameters, the fleet optimization solution
for the emission limit of 120 gCkmO2 is considered which is held constant for all
parameter variations. The sensitivity analysis is performed for the parameters of
measure availability, fuel prices in several scenarios and material prices. For the
sensitivity analysis, only financial parameters are varied as the environmental
results are assumed to be fixed. Consequently, the tool user has no influence on
the environmental data. In detail, the following parameter variations are analyzed:

• 100% availability of all measures and options

• All fuel and energy prices set to 115% of the normal case values
6.4 Results of Optimization for Different Scenarios 185

• All fuel and energy prices set to 85% of the normal case values
• The material prices for the car bodies set to 115% of the normal case values
• The material prices for the car bodies set to 85% of the normal case values
• Pro CNG & eGas scenario: CNG & eGas prices set to 85%, other fuels and
energy prices set to 115%
• Pro gasoline & diesel scenario: gasoline- & diesel-prices set to 85%, other
fuels and energy prices set to 115%
• Pro electrified scenario: electric energy prices set to 85%, other fuels and
energy prices set to 115%

Figure 6.23 shows how these parameter variations influence the result of the fleet
optimization in terms of a more cost-efficient or a more expensive average fleet
vehicle. The biggest influence on the average vehicle TCO is the variation of
all fuel and energy prices either by increasing or decreasing all prices by 15%.
This variation leads to either + 2.59% of the TCO to e56,815 per average fleet
vehicle (increased prices) or to -2.84% with e53,810 (decreased prices). Varying
the material prices for the car body (aluminum and steel costs) by ± 15% results
in rather small changes for the total TCO value with a change of ± 0.27% to
e55,531 (increased prices) and e55,231 (decreased prices).
Changing the availability of all possible options and measures to 100% in
order to let the algorithm freely chose the ideal fleet configuration leads to a
TCO value of only e54,958. Consequently, the parameter of module availability
influences the result by -0.76%. Varying the fuel and energy prices in differ-
ent manners enables the analysis of model sensitivity concerning opposite price
developments of fuels for different powertrain concepts. The biggest effect on
the total vehicle TCO is caused by setting the gasoline and diesel prices to 85%
of the original value and all other fuel costs to 115%. This price scenario leads
to a final TCO value of -1.04% to e54,807. A price development scenario pro
CNG and eGas results in a TCO value of + 0.17% to e55,476, a scenario pro
electrified powertrains results in additional costs of e245 to e55,626 (+0.44%).
Figure 6.24 shows the fleet composition in regard to powertrain shares for
each scenario of the sensitivity analysis. It can be seen that the variation of the
least sensitive parameter (material prices) does not lead to a change in the fleet
composition. Both scenarios (increased and decreased prices) have the same fleet
composition as the baseline standard scenario with the regular data base. This
is because the variation of car body prices influences all powertrain concepts
in the same way and does not lead to changed price differences between the
powertrain options that cause an altered fleet composition. A slight change in
the fleet configuration is reached by the reduction of all fuel and energy costs.
186 6 Application of the Optimization Approach to a Case Study …

fleet average: limit 120 g CO2

= 55,381 € (standard)

= 54,958 € (100 % all avail.)

emissions [g CO2 eq. / km]

= 53,810 € (all fuels -15 %)

= 56,815 € (all fuels + 15 %)

+2.59 %

+0.44 % = 55,531 € (materials +15 %)

+0.27 %
= 55,231 € (materials -15 %)
+0.17 %

-0.27 % = 54,807 € (pro gasoline & diesel)

-0.76 %
= 55,476 € (pro CNG & eGas)
-1.04 %

-2.84 % = 55,626 € (pro electrified)

total cost of ownership [€]

Figure 6.23 Results of sensitivity analysis

The new fleet design erases all CNG vehicles and replaces them by more diesel
driven vehicles (fourth bar). The parameter variations of increasing all fuel costs
by 15% and the price development scenarios pro CNG and eGas and also pro
electrified powertrains result in the same fleet composition with 38.2% of CNG
based vehicles, 34% of BEV configurations and between 6% to 8.4% for BEVs
with green electricity, eGas, diesel and PHEVs with green electricity. The price
variation to a pro gasoline and diesel scenario leads to a drastically changed share
of powertrains with 73.9% of diesel-based vehicles and minor shares of PHEVs,
BEVs and eGas vehicles. When setting all availabilities to 100%, the fleet design
is much more simplified. In this case, the ideal fleet consists of 42.6% of CNG
based vehicles and 57.4% of BEVs.
The results of the sensitivity analysis show that regarding the TCOs, the
biggest sensitivity exists in regard to the variation of all fuel and energy prices
by ± 15%. Regarding the fleet composition, the biggest sensitivities exist for
the variation of divergent price developments for fuel prices with a scenario pro
gasoline and diesel. Regarding the parameter of module availability, the biggest
sensitivities exist for the possible measures and options for vehicle configuration.
The results regarding the parameter sensitivity of the model can differ if a differ-
ent baseline scenario in terms of emission level is chosen as the fleet design and
TCO value also change with every change in the target emission level.
6.5 Findings for Further Vehicle Development and Fleet Planning 187

100 %
Share of powertrains

0%
Standard materials materials All fuels All fuels Pro CNG & Pro Pro gasoline 100% all
+15 % -15 % -15 % +15 % eGas electrified & diesel available

Diesel CNG PHEV BEV BEV Green CNG eGas

Green

Figure 6.24 Fleet composition for sensitivity analysis scenarios

6.5 Findings for Further Vehicle Development and Fleet

Planning

The analysis of the four selected scenarios shows the ideal vehicle or fleet con-
figuration for the given emission limitation. Figure 6.25 depicts the results of all
scenarios of the single vehicle and fleet optimization. These results show that for
the first scenario, both the single vehicle and fleet optimization results are based
on the same vehicle configuration. For this scenario, the differentiated optimiza-
tion of single vehicle and fleets does not lead to different results. For the next
two scenarios with stricter emission targets, the optimization on fleet level results
in an average TCO that is below the single vehicle value. In these cases, the
optimization on fleet level shows better results than the optimization (and scale
up to fleet level) of a single vehicle. For the last scenario, the results change, as
the single vehicle optimization has the lower TCO value than the optimized fleet.
This is due to the strongly limited vehicle availability for low emission vehicles.
 scenario 1
188

1A scenario 2
2

scenario 3
6

3
E1

4
scenario 4
E2

emissions [g CO2 eq. / km]

• Gasoline
• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

total cost of ownership [€]

difference to fleet soluon difference to fleet soluon single vehicle, 125 g CO2 / km: single vehicle, 105 g CO2 / km:

A = 0.0 g CO2 eq. | 0 € B = - 14.2 g CO2 eq. | + 74 € E1 = - 5.7 g CO2 eq.| + 492 € E2 = - 3.9 g CO2 eq. | - 2,249 €

Figure 6.25 Comparison of vehicle and fleet results

Application of the Optimization Approach to a Case Study …
6.5 Findings for Further Vehicle Development and Fleet Planning 189

The comparison of single vehicle optimization results with fleet results shows
that for some emission limits either the fleet solution or the single vehicle solution
shows better results. That means that a smart vehicle fleet design helps to reduce
costs. Additionally, it is possible that a single vehicle scope might lead to wrong
conclusions as scenario four (emission limit of 105 gCkmO2 ) shows. The single
vehicle result can mislead the user that relatively low TCO values are possible
for such low emission limits as the availability is not considered. Only the fleet
optimization shows that the average vehicle fleet (which is the focus for the
legislation) has much higher TCO values for the same limit. While the single
vehicle optimization is best applied by vehicle developers and technical engineers
/ experts, the fleet planning scope addresses the needs of decision makers for fleet
emission compliance.
The individual vehicle configurations that are part of the identified solutions
for fleet optimization show that most often the choice and shares of powertrain
is the most used approach to reach a given emission limit. The further surround-
ing measures such as weight reduction or alternative transportation etc. are only
integrated in the fleets if an emission limit cannot be reached anymore by only
adjusting the share of powertrains. A further aspect for the selection of vehi-
cles that are part of the final fleet solution is the approximation of the vehicle
towards the auxiliary line that marks the efficient frontier of eco-efficiency for
each emission level in the scatter plot.
The algorithm prefers to integrate vehicles in the fleets that are as close as
possible to this line regarding their emission and cost ratio. Except for the PHEV
vehicles, all powertrain groups have several vehicle configurations that show a
proximity to this line, making these powertrains preferred choices for the fleet
configuration. All PHEV vehicles are only part of the solution for the introduced
scenarios if no other option is available due to underlying restrictions and module
limitations.
Figure 6.26 shows an overview of fleet optimization solutions for 15 different
scenarios (ranging from 175 gCkmO2 to 105 gCkmO2 ).
This figure shows that for the first emission limits, an emission reduction with
relatively low extra costs can be achieved. For the emission limits of 120 gCkmO2
and lower, the extra costs grow quickly, making a reduction of 5 g of CO2 per
km much more expensive than in the beginning as all simpler measures have
already been integrated in the fleets. From 120 gCkmO2 and onwards, the line that
connects the individual fleet results starts to drift away from the efficient fron-
tier of the emission and cost ratio. This is an indicator for the decision maker
to not only invest in more expensive vehicle configurations, but also to invest
in infrastructure etc. to increase the availability of vehicle configurations with
 190
6

emissions [g CO2 eq. / km]

• Gasoline
• CNG
• Diesel
• PHEV
• BEV
• PHEV – Green Power
• CNG – eGas
• BEV – Green Power

total cost of ownership [€]

Figure 6.26 Result overview for all fleet scenarios

Application of the Optimization Approach to a Case Study …
6.5 Findings for Further Vehicle Development and Fleet Planning 191

a lower emission profile along the ideal auxiliary line. Considering such invest-
ments, the optimization algorithm is able to find new fleet compositions to meet
lower emission limits with more cost-efficient fleets.
Figure 6.27 shows the share of each powertrain type for the 15 different
emission level scenarios.

100 %
Share of powertrains

0%
175 g 170 g 165 g 160 g 155 g 150 g 145 g 140 g 135 g 130 g 125 g 120 g 115 g 110 g 105 g
CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2 CO2

Gasoline Diesel CNG PHEV PHEV BEV BEV Green CNG eGas
Green

Figure 6.27 Share of powertrains for all emission limit scenarios

The diagrams show that each powertrain type (disregarding the detailed vehi-
cle configurations) has an emission level on which it is more dominant than on
other emission levels. Vehicles equipped with a gasoline driven powertrain e.g.
are very dominant for the emission levels with relatively high limit values (fur-
ther to the left). In view of shrinking emission limit values, this dominance is
reduced, and the gasoline-based vehicles are replaced by more environmentally
friendly powertrains as the gasoline vehicle itself cannot reach the required levels
of emission anymore.
Figure 6.27 also shows that vehicles based on CNG and diesel have a broad
range of application for nearly all analyzed emission levels. This illustrates that
these powertrains have such a good ratio of emissions to costs with a relatively
low TCO value that the algorithm tries to include these powertrains as often as
possible into the fleets to keep the average TCO value as low as possible.
Another observation is that the CNG-based vehicles and the BEVs are flexible
regarding their emission value and are therefore often included in the final fleets.
192 6 Application of the Optimization Approach to a Case Study …

If CNG vehicles with regular CNG and CNG vehicles powered by eGas from
regenerative energy sources are combined, the technology of CNG based power-
trains offers a broad range of GHG emissions and TCOs. This range makes this
technology suitable for the flexible integration into fleets for different emission
limit values. The same applies for BEV vehicles which can either be powered by
the regular electricity mix or explicitly with renewable energy. This makes the
BEV very flexible regarding TCO and emission values and, thus, very suitable
for a flexible fleet planning as development costs for additional powertrains can
be reduced if such flexible powertrains are available.
Summary, Critical Appraisal
and Outlook 7

7.1 Summary

This work describes a methodology that helps to identify the ideal combination of
alternative components to create a modular product system with the aim to reach
a given environmental target with the lowest financial expenses over lifetime
and, hence, to increase the eco-efficiency. If transferred to the level of vehicle
fleets and absolute GHG limits for the entire fleet, the approach also supports the
concept of eco-effectiveness.
Due to an increasing relevance of compliance with environmental limitations,
the reduction of e.g. greenhouse gas emissions becomes an important target for
the producing industry. A prominent example is the automotive industry. In this
industry, a complex product system with a long supply chain and an energy-
intensive use stage meets a sophisticated but varying legislation for different
regions. While the overall limitation of greenhouse gases is set as a fleet aver-
age for automotive OEMs, the design of an individual vehicle within this fleet is
determined by the OEM. Herein lies a huge optimization potential, as an opti-
mized configuration of individual products and their integration into a fleet helps
to reach the emission limitation with a setup of different vehicles that result in
the lowest financial costs.
Given the opportunity to improve the eco-effectiveness, the current state of
academic knowledge does not provide a comprehensive solution as no streamlined
methodology exists that holistically address this problem. The established and
currently available approaches regarding the optimization of a product system’s
environmental impact lack different important aspects that enable a data-driven
decision support for OEMs. One important aspect is the handling of a quickly

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden 193

GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6_7
194 7 Summary, Critical Appraisal and Outlook

increasing number of possible alternative module combinations while being able

to ignore combinations that are either technically not feasible or unwanted due
to other reasons. Furthermore, it has to be considered that the inclusion of a
new module into the product system may cause interdependencies with another
module resulting in changed properties. These aspects also need to be transferred
from the optimization of a single product system to the requirements of a fleet
optimization. Most relevant is the consideration of all individual product sys-
tems, but also the combined fleet results. A complex challenge that arises with
the fleet perspective is the consideration of a limited availability of several com-
ponents. This requires assigning these limited options to those individual vehicles
within the fleet where this module unfolds its biggest potential due to positive
interdependencies with other modules.
To merge all these aspects into one methodology, a framework has been
developed that compromises data provision, data management, optimization, and
data visualization. Key elements are the optimization algorithm based on graph
theory and the design of the product system’s structure network. The product
system’s network provides the layout of all possible module combinations that
result in a final product and sets the solution space of the algorithm. The algo-
rithm is based on a minimum-cost flow problem from the field of graph theory
that is programmed to find the shortest path through the product system’s net-
work. The weightings of each arc in the network are based on the environmental
impact. Consequently, the shortest path results in the lowest emission profile. As
a result, the algorithm indicates the necessary modules in order to yield an over-
all emission average that meets the emission target while having the lowest total
costs.
Applied to the automotive industry, the methodology provides decision support
regarding the technical development of new product systems and the strategic
composition of a whole vehicle fleet. With regard to the support for technical
development, the algorithm helps to assign limited technologies or materials to
those vehicles where their impact is maximized by smart module interactions. In
terms of fleet optimization, it becomes clear that strategies with a direct focus on
fleets lead to better results regarding the eco-effectiveness than the optimization
of a single vehicle and the following scale up of that single solution to a fleet
level. Additional findings are the understanding of the best cost-to-emission ratio
of each powertrain. This helps to understand which powertrain dominates for
different emission targets. From that it can be derived which technology should
be the basis of a fleet for e.g. different world regions with different emission
limits. The results show that for low emission standards, the combustion-based
vehicles with a lightweight design or a lower rolling resistance are in favor while
7.2 Critical Appraisal 195

for stricter limitations electrified powertrains are dominant, given the introduced
scenarios.

7.2 Critical Appraisal

The introduced method has its strengths in finding the ideal combination of prod-
uct system configurations that together form a fleet or group of products that need
to meet a given average environmental target. Even with extensive numbers of
individual products within a fleet, the methodology is able to quickly find solu-
tions as the graph theory-based approach allows large data analysis in a short
time frame.
However, the developed concept requires several preconditions and available
input data to unfold its full potential. It is expected that the analyzed product
system is already well known and fully analyzed regarding its financial and envi-
ronmental performance. That means that a detailed LCC and LCA have been
performed and the results can be accessed in total or in parts. Also, it is required
to have all this data and information available at an early stage of the product
system’s development to incorporate the results of the optimization into the plan-
ning of a vehicle fleet composition. By increasing the quality of the input data,
the quality of the results and recommendations of the developed methodology
increase as well. Reducing uncertainties and data gaps is therefore a crucial suc-
cess factor. If important components of the product system or important parts
of the life cycle stages are not modelled and sufficiently calculated by LCC and
LCA tools, the derived results may mislead the user. Also, the results for the
LCCs and especially LCAs are considered to be fixed values that are absolutely
accurate. However, LCAs and LCCs have their own uncertainties which may
lead to varying results. When combining multiple modular LCAs where each
LCA has an own uncertainty level, the uncertainties multiply. This can finally
lead to a result for the full product system’s LCA which is far off from the real
value. The present method does not consider such uncertainty propagations in the
input values.
The results of this methodology are most interesting when applied to a com-
plex product system where a large solution space is created. This requires the
handling of huge amounts of input data that need to be structured in a data struc-
ture table and a network structure to model the product system. As the example
shown in this work is applied to a limited and simplified model of an automotive
life cycle, the results cannot be transferred directly to a real vehicle fleet design.
This network can be expanded to further production steps and more components
196 7 Summary, Critical Appraisal and Outlook

of a vehicle to make sure that no hidden hotspots are missed out to increase
the solution quality. Rising numbers of considered module categories and higher
number of alternatives per category lead to a higher complexity of managing such
data amounts with the introduced structure tables and structure networks.
The analyzed scenarios are designed based on the current emission targets.
However, with upcoming regulations even stricter emission levels are possible.
Such moving targets for GHG emission reductions lead to a situation where the
current targets are not ambitious enough. In such a case, scenarios should be
analyzed that directly anticipate further emission reductions. Those scenarios will
then most likely recommend even higher shares of alternative powertrains, such
as BEVs, than the current scenarios. A higher share of e.g. BEVs in turn leads to
reduced costs as the development costs can be distributed over a larger number
of vehicles. Such economies of scale are currently not considered in the model.
Each step of the framework introduced in section 4.1 requires the utilization of
software support to either manage the data, to calculate and provide input data for
the LCC and LCA, to optimize this data input and to interpret and visualize the
optimization results. The way the developed methodology is implemented (see
chapter 5), different software applications are required for each step. Separating
the steps of the methodology on different software systems leads to additional
effort of transferring the data into and out of each software. Due to the prototyp-
ical implementation, the imports and exports of data between different software
need to be performed partially by manual operations. The manual data transfer
has two disadvantages: (1) it takes more time to set up the whole methodology
and (2) it is more vulnerable to mistakes. A more accessible software environment
can be created as described in the following outlook to overcome this.

7.3 Outlook

Further research in this field should focus on (1) improving the presented method-
ology in short and long-term perspectives and on (2) transferring the approach or
the findings to other fields of application.

Improvement of methodology
The introduced approach can be improved by further research in order to optimize
the solution quality, the considered information variety or the usability of the tool.
The current approach of the developed methodology is focused on the given
boundary conditions and the fixed solution space of which the algorithm identifies
the combination that comes the closest to the defined optimization targets. This
7.3 Outlook 197

makes the approach limited to the initially defined “reality” from which it cannot
break out. A defined availability of e.g. green electricity cannot be increased during
a scenario. The availability of different modules can be given more degrees of free-
dom by making it possible to increase limited modules by an additional investment
condition to bring the model closer to the possibilities of the reality. Additional
nodes for e.g. further wind power turbines can be added which can be chosen by the
algorithm. The choice of an additional wind turbine requires high investment costs
and the additional green electricity can then be used to power a defined amount of
additional BEV. In this way, the algorithm starts with an initial infrastructure level,
but would result in an overall ideal situation regarding the availability of limited
modules, as this concept can also be applied to additional capacities for e.g. eGas
or aluminum car bodies.
Another form of expanding the methodology in terms of a broader information
basis is the incorporation of further environmental midpoint impact categories such
as e.g. the eutrophication potential, acidification potential or resource depletion
potential. By this it can be avoided that the calculated result is beneficial with
regard to one environmental aspect (e.g. climate change), but has negative effects
for other categories. This can prevent problem shifting from one environmental
category to another. However, a switch to a multi-criteria optimization is required
to incorporate all categories in one solution. Furthermore, a way of including the
uncertainty propagation in the input data can also strengthen the result quality.
The input data for the optimization algorithm for the environmental impact is
provided by attributional LCAs to quantify the allocated shares of GHG emissions
for each component or module by applying normative cut-off approaches from
surrounding influences outside the own product system. The environmental impacts
can also be assessed by consequential LCAs in further studies to fully assess the
impact of a selected solution for a product fleet. Those consequential LCAs reflect on
the effects that certain solutions (e.g. different types of powertrains for vehicles) will
have on surrounding ecosystems. By including consequential LCAs, the holistic and
predictive aspect of the presented approach for product fleets can be strengthened
even further. Social aspects could also be incorporated to cover all three pillars of
sustainable development.
The current financial perspective is focused on an overall life cycle costing with
the aim to minimize costs as a society. From an OEM’s perspective, a switch of
this perspective from minimizing costs to maximizing profits can be an additionally
interesting result. In such a scenario, the optimization target still is to meet the
emission targets, but instead of reducing the total cost of ownership, the profit
margin of an OEM can be maximized. Also relevant for an OEM is the extension
of considering multiple vehicle types in one optimization. Such an extension can
198 7 Summary, Critical Appraisal and Outlook

help to answer the question “How many SUVs can be sold while still meeting the
emission limitation?”. While the current methodology results in lowest costs and
recommends e.g. a fleet of 100% small vehicles, an algorithm with the target of
profit maximization finds the ideal balance between expensive cars with a higher
profit margin and low emission vehicles to make sure the average emission target is
met.
As introduced in chapter 5, the presented methodology is designed as an indepen-
dent tool with interfaces for input and output to other software solutions for e.g. the
LCA calculation or the calculation of the result by the optimization algorithm. The
transfer of information between the separated systems has several disadvantages as
described in the critical appraisal. Either a new and comprehensive software suite has
to be programed in a way that includes all required tasks (provision and calculation
of input data, data management, optimization, and visualization) in one application
or a supportive tool has to be developed that automizes the data transfer between
the different systems to overcome this. A further option is to include the developed
approach of this work into existing softwares for LCA or LCC calculation.

Transfer to other applications

The developed methodology can also be applied to use cases outside of the pre-
sented application for the automotive industry. Industries that could profit from
such an optimization approach include ship construction, airplane construction,
mobile phone industry, or housing and building construction. Wherever a product
system is designed that offers different modular components and is used in a long
and energy-intensive use stage, a holistic optimization can help to improve the over-
all environmental performance. Furthermore, the planning of a multi-modal public
transportation system for cities and urban areas can be considered. In such a scenario,
there are different alternatives for the provision of mobility that offer different func-
tionalities with different environmental and financial impacts. In such a scenario,
the algorithm could choose between e.g. a subway network, an overground tram,
smaller and bigger busses, taxis, car sharing, ride hailing and other options. For
decision makers in urban areas, it is highly relevant to offer mobility services that
fulfill all demands while reducing environmental impacts like land use, air pollution,
noise pollution and GHG emissions. The given methodology assists in identifying
an ideal mix of all possible mobility options that offer mobility for a defined num-
ber of people within a defined area while meeting an exemplary city-wide emission
limit.
Appendix

© The Editor(s) (if applicable) and The Author(s), under exclusive license 199
to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6
 200

Table A.1 Data structure table with algorithm input data

Appendix
 Appendix

Table A.1 (continued)

201
 202

Table A.1 (continued)

Appendix
Literature

ACEA (2020): Facts about the Automobile Industry, URL: https://www.acea.be/automobile-

industry/facts-about-the-industry, last checked on 28.07.2020.
ADAC (2019a): ADAC Autokosten, VW Golf VII in various powertrains, URL: https://www.
adac.de/infotestrat/autodatenbank/autokosten/default.aspx, last checked on 11.02.2019.
ADAC (2019b): VW e-Golf (04/17 – 05/20), URL: https://www.adac.de/infotestrat/tests/auto-
test/detail.aspx?IDTest=5678, last checked on 20.02.2019.
ADAC (2019c): ADAC Auto Test VW Golf GTE DSG, URL: https://www.adac.de/_ext/itr/
tests/autotest/AT5233_VW_Golf_GTE_DSG/VW_Golf_GTE_DSG.pdf, last checked on
20.02.2019.
Alting, L. (1995): Life Cycle Engineering and Design. In: CIRP Annals — Manufacturing
Technology 44 (1995), No. 2, p. 569–580.
Ameli, M.; Mansour, S.; Ahmadi-Javid, A. (2017): A sustainable method for optimizing
product design with trade-off between life cycle cost and environmental impact. In: Envi-
ronment, Development and Sustainability 19 (2017), p. 2443–2456.
Azapagic, A.; Clift, R. (1998): Linear Programming as a Tool in Life Cycle Assessment. In:
International Journal of Life Cycle Assessment 3 (1998), No. 6, p. 305–316.
Azapagic, A.; Clift, R. (1999): Life cycle assessment and multiobjective optimization. In:
Journal of Cleaner Production 7 (1999), p. 135–143.
Bataglin, M.; Ferreira, J. C E. (2020): A modularization method based on the triple bottom
line and product desirability: A case study of a hydraulic product. In: Journal of Cleaner
Production 271 (2020), p. 1–15.
Becker, J. (2019): Elektro-Strategie von Volkswagen, Der Riese erwacht. In: Süddeutsche
Zeitung GmbH (18.02.2019), URL: https://www.sueddeutsche.de/auto/vw-tesla-elektr
oautos-1.4326717, 18.02.2019, last checked on 20.02.2019.
Braungart, M.; McDonough, W.; Bollinger, A. (2007): Cradle-to-cradle design: creating
healthy emissions e a strategy for eco-effective product and system design. In: Journal
of Cleaner Production 15 (2007), p. 1337–1348.
Broch, F. (2017): Integration von ökologischen Lebenswegbewertungen in Fahrzeugentwick-
lungsprozesse. Wiesbaden: Springer, 2017.
Buchert, T.; Neugebauer, S.; Schenker, S.; Lindow, K.; Stark, R. (2015): Multi-criteria deci-
sion making as a tool for sustainable product development—Benefits and obstacles. In:
Procedia CIRP 26 (2015), p. 70–75.

© The Editor(s) (if applicable) and The Author(s), under exclusive license 203
to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2022
C. D. Gabrisch, Eco-Effectiveness of Modular Products and Fleets within the Auto-
motive Industry, AutoUni – Schriftenreihe 164,
https://doi.org/10.1007/978-3-658-40594-6
204 Literature

Bundesnetzagentur; Bundeskartellamt (2021): Monitoringbericht 2021. Bonn: MKL Druck

GmbH & Co. KG, 2021.
Buxmann, K.; Kistler, P.; Rebitzer, G. (2009): Independent information modules—a power-
ful approach for life cycle management. In: International Journal of Life Cycle Assess-
ment 14 (2009), p. 92–100.
Chen, W.; Rosen, D.; Allen, J.; Mistree, F. (1994): Modularity and the Independence of Func-
tional Requirements in Designing Complex Systems. In: Concurrent Product Design 74
(1994), p. 31–38.
Chiu, M.-C.; Chu, C.-H. (2012): Review of Sustainable Product Design from Life Cycle
Perspectives. In: International Journal of Precision Engineering and Manufacturing 13
(2012), p. 1259–1272.
Chivers, I.; Sleightholme, J. (2018): Introduction to Programming with Fortran. 4th Edition.
Cham: Springer International, 2018.
Cerri, D.; Taisch, M.; Terzi, S. (2014): Proposal of a model for Life Cycle Optimization of
industrial equipment. In: 21st CIRP Conference on Life Cycle Engineering 15 (2014),
p. 479–483.
Dahmus, J.; Gonzalez-Zugasti, J.; Otto, K. (2001): Modular product architecture. In: Design
Studies 22 (2001), No. 5, p. 409–424.
Deutscher Bundestag (1998): Konzept Nachhaltigkeit—Vom Leitbild zur Umsetzung.
Schutz des Menschen und der Umwelt—Ziele und Rahmenbedingungen einer nachhaltig
zukunftsverträglichen Entwicklung. In: UWSF-Z. Uweltchem. Okötex 10 (1998), No. 244,
p. 1–252.
DIN ISO 14040: Environmental Management—Life Cycle Assessment—Principles and
Framework. European Standard. ISO, Geneva, 2006.
DIN ISO 14044: Environmental Management—Life Cycle Assessment—Requirements and
Guidelines. European Standard. ISO, Geneva, 2006.
Domschke, W.; Drexl, A.; Klein, R.; Scholl, A. (2015): Einführung in Operations Research.
9th Edition. Berlin, Heidelberg: Springer Verlag, 2015.
Dose, J. (2005): Austauschbare Sachbilanzmodule auf Basis von Produktkomponenten.
Technische Universität Berlin, Dissertation, 2005.
Dyckhoff, H.; Spengler, T. (2007): Produktionswirtschaft. Eine Einführung für Wirtschaftsin-
genieure. 2nd Edition. Berlin, Heidelberg: Springer-Verlag, 2007.
Ehrlich, P.; Holdren, J. (1971). Impact of Population Growth. In: Science 171 (1971), No.
3977, p. 1212–1217.
Eichhorn, P.; Merk, J. (2016): Das Prinzip Wirtschaftlichkeit. Basiswissen der Betrieb-
swirtschaftslehre. 4th Edition. Wiesbaden: Springer Gabler Verlag, 2016.
Elkington, J. (2018): Recalling the triple bottom line, URL: https://johnelkington.com/
2018/07/recalling-the-triple-bottom-line/, Zen and the triple bottom line, last checked on
05.07.2020.
Ellinger, T.; Beuermann, G.; Leisten, R. (2003): Operations Research. Eine Einführung. 6th
Edition. Berlin, Heidelberg: Springer, 2003.
ElMaraghy, H.; AlGeddawy, T. (2013): Optimum granularity level of modular product design
architecture. In: CIRP Annals—Manufacturing Technology 62 (2013), No. 1, p. 151–154.
Ercan, T.; Zhao, Y.; Tatari, O.; Pazour, J. (2015): Optimization of transit bus fleet’s life cycle
assessment impacts with alternative fuel options. In: Energy 93 (2015), No. 1, p. 323–334.
Literature 205

Ernst, H.; Schmidt, J.; Beneken, G. H. (2016): Grundkurs Informatik Grundlagen und
Konzepte für die erfolgreiche IT-Praxis—Eine umfassende, praxisorientierte Einführung.
6th Edition. Wiesbaden: Springer Vieweg, 2016.
European Commission (2015): EU ETS Handbook, URL: https://ec.europa.eu/clima/sites/
clima/files/docs/ets_handbook_en.pdf, last checked on 04.02.2020
European Commission (2019): The European Green Deal, URL: https://ec.europa.eu/
info/sites/default/files/european-green-deal-communication_en.pdf, last checked on:
17.10.2020.
European Commission (2020a): Reducing CO2 emissions from passenger cars—before
2020, URL: https://ec.europa.eu/clima/policies/transport/vehicles/cars_en, last checked
on 16.06.2020.
European Commission (2020b): CO2 emission performance standards for cars and
vans (2020 onwards), URL: https://ec.europa.eu/clima/policies/transport/vehicles/regula
tion_en, last checked on 16.06.2020.
European Commission (2020c): Stepping up Europe’s 2030 climate ambition, Investing in
a climate-neutral future for the benefit of our people, URL: https://eur-lex.europa.eu/
legal-content/EN/TXT/PDF/?uri=CELEX:52020cDC0562&from=EN, last checked on
08.09.2021.
European Parliament (2000): Directive 2000/53/EC of the European Parliament of 18 Septem-
ber 2000 on end of life vehicles, URL: https://eur-lex.europa.eu/resource.html?uri=cellar:
02fa83cf-bf28-4afc-8f9f-eb201bd61813.0005.02/DOC_1&format=PDF, last checked on
13.09.2019.
European Parliament (2007): Regulation 715/2007 of 20 June 2007 on type approval
of motor vehicles with respect to emissions from light passenger and commer-
cial vehicles (Euro 5 and Euro 6) and on access to vehicle repair and main-
tenance information, URL: https://eur-lex.europa.eu/legalcontent/EN/TXT/PDF/?uri=
CELEX:32007R0715, last checked on 13.09.2019.
European Parliament (2009): Regulation 443/2009 of the European Parliament and the Coun-
cil of 23 April 2009 setting emission performance standards for new passenger cars
as part of the Community’s integrated approach to reduce CO2 emissions from light-
duty vehicles, URL: https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:
32009R0443, last checked on 13.09.2019.
European Parliament (2013): Decision No 1386/2013/EU of the European Parliament and
the Council of 20 November 2013 on a General Union Environment Action Programme to
2020 ‘Living well, within the limits of our planet’, URL: https://eur-lex.europa.eu/legal-
content/EN/TXT/PDF/?uri=CELEX:32013D1386, last checked on 13.09.2019.
European Parliament (2014): Regulation No. 333/2014 of the European Parliament and
the Council of 11 March 2014 amending Regulation (EC) No 443/2009 to define the
modalities for reaching the 2020 target to reduce CO2 emissions from new passen-
ger cars, URL: https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:320
14R0333, last checked on 13.09.2019.
European Parliament (2019): Regulation 2019/631 of the European Parliament and the Coun-
cil of 17 April 2019 setting CO2 emission performance standards for new passenger cars
and for new light commercial vehicles, and repealing Regulations (EC) No 443/2009
and (EU) No 510/2011, URL: https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=
CELEX:32019R0631, last checked on 04.02.2020.
206 Literature

European Parliament (2020): CO2 emissions from cars: facts and figures, URL: https://
www.europarl.europa.eu/news/en/headlines/society/20190313STO31218/co2-emissi
ons-from-cars-facts-and-figures-infographics, last checked on 02.03.2020.
Feess, E. (2018): Gabler Wirtschaftslexikon—Definition “Umweltziele”, URL: https://wir
tschaftslexikon.gabler.de/definition/umweltziele-49477/version-126946, last checked on
12.11.2018.
FME (Federal Ministry for the Environment, Nature Conservation, Building and Nuclear
Safety) (2016): German Climate Action Plan 2050. Principles and goals of the German
government’s climate policy, Berlin: Druck- und Verlagshaus Zarbock, 2016.
Gabrisch, C.; Cerdas, F.; Hermann, C. (2019): Product System Modularization in LCA
Towards a Graph Theory Based Optimization for Product Design Alternatives. In:
Progress in Life Cycle Assessment, Sustainable Production, Life Cycle Engineering and
Management (2019), p. 37–44.
Garcia, R.; Gregory, J.; Freire, F. (2015): Dynamic fleet-based life-cycle greenhouse gas
assessment of the introduction of electric vehicles in the Portuguese light-duty fleet. In:
International Journal of Life Cycle Assessment 20 (2015), p. 1287–1299.
Gritzmann, P. (2013): Grundlagen der Mathematischen Optimierung. Diskrete Strukturen,
Komplexitätstheorie, Konvexitätstheorie, Lineare Optimierung, Simplex-Algorithmus,
Dualität. Wiesbaden: Springer Fachmedien, 2013.
Gu, P.; Sosale, S. (1999): Product modularization for life cycle engineering. In: Robotics and
computer integrated manufacturing 15 (1999), No. 5, p. 387–401.
Günther, E. (2018): Gabler Wirtschaftslexikon—Definition “Natürliche Umwelt”, URL:
https://wirtschaftslexikon.gabler.de/definition/natuerliche-umwelt-41343/version-
264708, last checked on 12.11.2018.
Halstenberg, F.; Buchert, T.; Bonvoisin, J.; Lindow, K.; Stark, R. (2015): Target-oriented
Modularization–Addressing Sustainability Design Goals in Product Modularization. In:
Procedia CIRP 29 (2015), p. 603–608.
Harvey, D. (2018): Cost and energy performance of advanced light duty vehicles: Implica-
tions for standards and subsidies. In: Energy Policy 114 (2018), p. 1–12.
Hauschild, M. (2015): Better—but is it good enough? On the need to consider both eco-
efficiency and eco-effectiveness to gauge industrial sustainability. The 22nd CIRP con-
ference on Life Cycle Engineering 29 (2015), p. 1–7.
Hauschild, M.; Huijbregts, M. (2015): Introducing Life Cycle Impact Assessment. In:
Hauschild, Michel Z.; Huijbregts, Mark A.J. Life Cycle Impact Assessment. Heidelberg,
New York, London: Springer Science+Business, 2015, p. 1–16.
Hauschild, M.; Herrmann, C.; Kara, S. (2017): An Integrated Framework for Life Cycle
Engineering. In: Procedia CIRP 61 (2017), p. 2–9.
Hauschild, M.; Kara, S.; Røpke, I. (2020): Absolute sustainability: Challenges to life cycle
engineering. In: CIRP Annals—Manufacturing Technology 00 (2020), p. 1–21.
Herrmann, C. (2010): Ganzheitliches Life Cycle Management. Nachhaltigkeit und Lebenszyk-
lusorientierung in Unternehmen. Berlin: Springer-Verlag, 2010.
Herrmann, I.; Lundberg-Jensen, M.; Jørgensen, A.; Stidsen, T.; Spliid, H.; Hauschild, M.
(2014): Enabling optimization in LCA: from “ad hoc” to “structural” LCA approach—
based on a biodiesel well-to-wheel case study. In: The International Journal of Life Cycle
Assessment 19 (2014), p. 194–205.
Literature 207

Hoch, G.; Heupel, T.; Kachel, T. (2016): Life-Cycle-Costing in der Unternehmenspraxis:

Techniken, Strategische Bedeutung, Umsetzungsprobleme. In: Becker, Wolfgang; Ulrich,
Patrick Handbuch Controlling. Wiesbaden: Springer Fachmedien, 2016, p. 329–344.
Hofstätter, H. (1977): Die Erfassung der langfristigen Absatzmöglichkeiten mit Hilfe des
Lebenszyklus eines Produktes. Modernes Marketing. Würzburg: Physica-Verlag, 1977.
ICCT (International Council on Clean Transportation) (2017): 2020–2030 CO2 stan-
dards for new cars and light-commercial vehicles in the European Union, 2017,
URL: https://theicct.org/sites/default/files/publications/ICCT_Post-2020-CO2-stds-EU_
briefing_20171026_rev20171129.pdf, last checked on 14.06.2020.
ICCT (International Council on Clean Transportation) (2020): Chart library: Passenger vehi-
cle fuel economy, 2020, URL https://theicct.org/chart-library-passenger-vehicle-fuel-eco
nomy, last checked on 14.06.2020.
IEA (2016): World Energy Outlook 2016, Paris: OECD/IEA, 2016.
IPCC (Intergovernmental Panel on Climate Change) (2014): Climate Change 2014: Miti-
gation of Climate Change. Contribution of Working Group III to the Fifth Assessment
Report of the Intergovernmental Panel on Climate Change [Edenhofer, O., R. Pichs-
Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner,
P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel and
J.C. Minx (eds.)]. Cambridge, United Kingdom and New York, NY, USA: Cambridge
University Press, 2014.
Islam, H.; Jollands, M.; Setunge, S.; Bhuiyan, M. (2015): Optimization approach of balanc-
ing life cycle cost and environmental impacts on residential building design. In: Energy
and Buildings 87 (2015), p. 282–292.
Jahandideh, A.; Aminikhangahi, S.; Salehnia, A.; Muthukumarappan, K. (2015): Life Cycle
Assessment and Sustainability Analysis of Lignin Derivative Products, Materials and
Technologies. Integrated Process Modeling, Scientific framework and LCA for Assess-
ing Sustainability. In: Conference American Society for Engineering Education, ASEE
Zone III, 2015.
Jakobsen, M. (1999): The Relation of Eco-Effectiveness and Eco-Efficiency—An Important
Goal in Design for Environment. 10. Symposium “Fertigungsgerechtes Konstruieren”,
Schnaittach, 14th and 15th October 1999.
Jensen, A.; Remmen, A. (2006): Background Report for a UNEP Guide to LIFE CYCLE
MANAGEMENT—A bridge to sustainable products. Paris: Division of Technology, Indus-
try and Economics, United Nations Environment Programme, 2006.
Jiao, J.; Simpson, T. W.; Siddique, Z. (2007): Product family design and platform-based
product development: a state-of-the-art review. In: J Intell Manuf 18 (2007), p. 5–29.
Jolliet, O.; Müller-Wenk, R.; Bare, J.; Brent, A.; Goedkoop, M.; Heijungs, R.; Itsubo, N.;
Peña, C.; Pennington, D.; Potting, J.; Rebitzer, G.; Stewart, M.; Haes, H. U.; Weideman,
B. (2004): The LCIA Midpoint-damage Framework of the UNEP/SETAC Life Cycle
Initiative. In: The International Journal of Life Cycle Assessment 9 (2004), No. 6,
p. 394–404.
Jose, A.; Tollenaere, M. (2005): Modular and platform methods for product family design:
literature analysis. In: Journal of Intelligent Manufacturing 16 (2005), p. 371–390.
Jungbluth, N. (2000): Umweltfolgen des Nahrungsmittelkonsums: Beurteilung von Produk-
tmerkmalen auf Grundlage einer modularen Ökobilanz. Technische Universität Berlin,
Dissertation, 2000.
208 Literature

Jungbluth, N.; Tietje, O.; Scholz, R. (2000): Food Purchases: Impacts from the customers
point of view investigated with a modular LCA. In: International Journal of Life Cycle
Assessment 5 (2000), No. 3, p. 134–142.
Kaltschmitt, M.; Schebek, L. (2015): Umweltbewertung für Ingenieure—Methoden und
Verfahren. Berlin, Heidelberg: Springer Vieweg, 2015.
Kaluza, A.; Kleemann, S.; Fröhlich, T.; Herrmann, C.; Vietor, T. (2017): Concurrent
design & life cycle engineering in automotive lightweight component development. In:
Procedia CIRP 66 (2017), p. 16–21.
Kara, S.; Hauschild, M.; Herrmann, C. (2018): Target-Driven Life Cycle Engineering: Stay-
ing within the Planetary Boundaries. In: Procedia CIRP 69 (2018), p. 3–10.
Kastner, W.; Schildt, G. (2005): Informatik. Aufgaben und Lösungen. Begleitbuch zu
Blieberger et al.: Informatik. Wien, New York: Springer, 2005.
Kaya, Y.; Yokobori, K. (1997): Environment, Energy, and Economy: strategies for sustain-
ability. United Nations Univ. Press, 1997.
Keoleian, G.; Kar, K.; Manion, M.; Bulkley, J. (1997): Industrial Ecology of the Automobile:
A Life Cycle Perspective. Warrendale, Pennsylvania: Society of Automotive Engineers,
International, 1997.
Kim, S.; Moon, S. K. (2019): Eco-modular product architecture identification and assessment
for product recovery. In: Journal of Intelligent Manufacturing 30 (2019), p. 383–403.
Klenter, G. (1995): Zeit—Strategischer Erfolgsfaktor von Industrieunternehmen. Dis-
sertation, Gerhard-Mercator Universität Duisburg, Duisburger Betriebswirtschaftlicher
Schriften, 1995.
Kloepffer, W.; Grahl, B. (2009): Ökobilanz (LCA) Ein Leitfaden für Ausbildung und Beruf .
Weinheim: Wiley VCH, 2009.
Kloepffer, W.; Ciroth, A. (2011): Is LCC relevant in a sustainability assessment? In: Inter-
national Journal of Life Cycle Assessment 16 (2011), p. 99–101.
Koeijer, B. B.; Wver, R.; Henseler, J. (2017): Realizing Product-Packaging Combinations in
Circular Systems: Shaping the Research Agenda. In: Packaging Technology and Science
30 (2017), p. 443–460.
Koffler, C. (2007): Automobile Produkt-Ökobilanzierung. Technische Universität Darmstadt,
Dissertation, 2007.
Koffler, C.; Rohde-Brandenburger, K. (2010): On the calculation of fuel savings through
lightweight design in automotive life cycle assessments, In: International Journal of Life
Cycle Assessment 15 (2010), p. 128–135.
Korte, B.; Vygen, J. (2018): Kombinatorische Optimierung—Theorie und Algorithmen. 3rd
Edition. Berlin: Springer-Verlag, 2018.
Kuo, T.-C.; Smith, S.; Smith, G.; Huang, S. (2016): A predictive product attribute driven eco-
design process using depth-first search. In: Journal of Cleaner Production 112 (2016),
p. 3201–3210.
Lemme, R.; Arruda, E.; Bahiense, L. (2019): Optimization model to assess electric vehicles
as an alternative for fleet composition in station-based car sharing systems. In: Trans-
portation Research 67 (2019), p. 173–196.
Mateika, M. (2005): Unterstützung der lebenszyklusorientierten Produktplanung am Beispiel
des Maschinen- und Anlagenbaus. Technische Universität Braunschweig, Dissertation,
2005.
Literature 209

Matiz-Rubio, N.; Eltrop, L.; Härdtlein, M. (2020): Nachhaltigkeitsbewertung bioökonomis-

cher Produktsysteme. In: Wilfried, Konrad; Scheer, Dirk; Weidtmann, Anette (eds.)
Bioökonomie nachhaltig gestalten. Perspektiven für ein zukunftsfähiges Wirtschaften.
Wiesbaden: Springer VS, 2020, p. 223–256.
Miah, J.; Koh, S. C. L.; Stone, D. (2017): A hybridised framework combining integrated
methods for environmental Life Cycle Assessment and Life Cycle Costing. In: Journal
of Cleaner Production 168 (2017), p. 846–866.
MoJ (Federal Ministry of Justice and Consumer Protection) (2020a): Gesetz zum Schutz
vor schädlichen Umwelteinwirkungen durch Luftverunreinigungen, Geräusche, Erschüt-
terungen und ähnliche Vorgänge (Bundes-Immissionsschutzgesetz—BimSchG), 2020a,
URL: https://www.gesetze-im-internet.de/bimschg/BImSchG.pdf, last checked on
08.02.2020a.
MoJ (Federal Ministry of Justice and Consumer Protection) (2020b): Gesetz über Abgaben
für das Einleiten von Abwasser in Gewässer (Abwasserabgabengesetz—AbwAG),
2020b, URL: http://www.gesetze-im-internet.de/abwag/AbwAG.pdf, last checked on
08.02.2020b.
MoJ (Federal Ministry of Justice and Consumer Protection) (2020c): Gesetz zum Schutz vor
gefährlichen Stoffen (Chemikaliengesetz—ChemG), 2020c, URL: https://www.gesetze-
im-internet.de/chemg/ChemG.pdf, last checked on 08.02.2020c.
MoJ (Federal Ministry of Justice and Consumer Protection) (2020d): Basic Law for the Fed-
eral Republic of Germany, Article 20a—Protection of the natural foundations of life and
animals, 2020, URL: https://www.gesetze-im-internet.de/gg/BJNR000010949.html, last
checked on 23.08.2020.
Müller, D. (2013): Betriebswirtschaftslehre für Ingenieure. Berlin, Heidelberg: Springer,
2013.
Mutingi, M.; Dube, P.; Mbohwa, C. (2017): A modular product design approach for sus-
tainable manufacturing in a fuzzy environment. In: Procedia Manufacturing 8 (2017),
p. 471–478.
Nadoveza D.; Koukias, A.; Karakoyun, F.; Kiritsis, D. (2013): A Concept for Graph-Based
LCA Analysis Tool. In: Prabhu, Vittal; Taisch Marco; Kiritsis, Dimitris (eds) Advances
in Production Management Systems. Sustainable Production and Service Supply Chains.
Heidelberg: Springer, 2013, p. 410–417.
Nickenig, K. (2016): Fachbegriffe Rechnungswesen und Steuerrecht Kaufmännisches Grund-
vokabular zum schnellen Nachschlagen für Praktiker und Lernende. Wiesbaden: Springer
Fachmedien, 2016.
Onat, N.; Kucuvar, M.; Tatari, O.; Zhen, Q. (2015): Combined application of multi-criteria
optimization and life-cycle sustainability assessment for optimal distribution of alterna-
tive passenger cars in U.S. In: Journal of Cleaner Production 112 (2015), p. 291–307.
O’Reilly, C.; Göransson, P.; Funazaki, A.; Suzuki, T.; Edlund, G.; Lundow, J.-O.; Cerin,
P.; Cameron, C., Wennhage, P.; Potting, J. (2016): Life cycle energy optimisation: A
proposed methodology for integrating environmental considerations early in the vehicle
engineering design process. In: Journal of Cleaner Production 135 (2016), p. 750–759.
Otto, H.; Mueller, K.; Kimura, F.; Germani, M.; Mandorli, F. (2002): Integrating life cycle
aspects within product family design: An example for SMEs. In: Kovacs, George; Bertok,
Peter; Haidegger, Geza (Eds.) Digital Enterprise Challenges. IFIP—the International
Federation for Information Processing, vol. 77. Boston: Springer, 2002.
210 Literature

Paasch, A.; Mueller, A. (2013): Vom Erz zum Auto—Rohstoffe für die Armen, schlechte
Arbeitsbedingungen für die Armen. Frankfurt/Main: Misereor—Welt-sichten, 2013.
Pandremenos, J.; Paralikas, J; Salonitis, K.; Chryssolouris, G. (2009): Modularity concepts
for the automotive industry: A critical review. In: CIRP Journal of Manufacturing Science
and Technology 1 (2009), p. 148–152.
Papageorgiou, M.; Leibold, M.; Buss, M. (2015): Optimierung. Statische, dynamische,
stochastische Verfahren für die Anwendung. 4th Edition. Berlin, Heidelberg: Springer,
2015.
Paul, J. (2015): Praxisorientierte Einführung in die Allgemeine Betriebswirtschaftslehre. 3rd
Edition. Wiesbaden: Springer Gabler Verlag, 2015.
Pfeiffer, W.; Bischof, P. (1981): Produktlebenszyklus—Instrument jeder strategischen Pro-
duktplanung. In: Steinmann, Horst (eds) Planung und Kontrolle—Probleme der strate-
gischen Unternehmensführung. München: Verlag Fran Vahlen, 1981, S. 133–166.
Plinke, W.; Rese, M.; Utzig, P. (2015): Industrielle Kostenrechnung. 8th Edition. Berlin:
Springer Vieweg, 2015.
Rebitzer, G. (2005): Enhancing the application efficiency of life cycle assessment for
industrial uses. EPFL Lausanne, Dissertation, 2005.
Rebitzer, G.; Hunkeler, D. (2003): Life Cycle Costing in LCM: Ambitions, Opportunities,
and Limitations. In: International Journal of Life Cycle Assessment 8 (2003), No. 5,
p. 253–256.
Reccioni, M.; Mandorli, F.; Germani, M.; Faraldi, P.; Polverini, D. (2007): Life-Cycle
Assessment simplification for modular products. In: Takata, Shozo; Umeda, Yasushi (eds)
Advances in Life Cycle Engineering for Sustainable Manufacturing Businesses. London:
Springer, 2007, p. 53–58.
Reichmuth, D.; Lutz, A.; Manley, D.; Keller, J. (2013): Comparison of the technical potential
for hydrogen, battery electric, and conventional light-duty vehicles to reduce greenhouse
gas emissions and petroleum consumption in the United States. In: International Journal
of Hydrogen Energy 38 (2013), p. 1200–1208.
Rockström, J.; Steffen, W.; Noone, K.; Persson, Å.; Folke, C.; Nykvist, B.; Sörlin, S.;
Costanza, R.; Svedin, U.; Falkenmark, M.; Karlberg L.; Walker, B. (2009): A safe oper-
ating space for humanity. In: Nature 461 (2009), p. 472–475.
Rohde-Brandenburger, K. (2013): Leichtbau—Was bringen 100 kg Gewichtsreduzierung im
Verbrauch? Eine physikalische Berechnung. In: Automobiltechnische Zeitschrift 115,
(2013), No. 7, p. 584–591.
Rubinstein-Salzedo, S. (2018): Cryptography. Cham: Springer International Publishing AG,
2018.
Rudorfer, M. (2017): Intensivkurs Kostenrechnung—Anschaulicher Einstieg für Studium und
Praxis. 2nd Edition. Wiesbaden: Springer Fachmedien, 2017.
Schaltegger, S.; Sturm, A. (1990): Ökologische Rationalität. Ansatzpunkte zur Aus-
gestaltung von ökologieorientierten Managementinstrumenten. In: Die Unternehmung.
Schweizerische Zeitschrift für betriebswirtschaftliche Forschung und Praxis 44 (1990),
No. 4, p. 273–290.
Schlink, H. (2014): Wirtschaftlichkeitsrechnung für Ingenieure Grundlagen für die Entwick-
lung technischer Produkte. Wiesbaden: Springer Fachmedien, 2014.
Literature 211

Schrack, D. (2016): Nachhaltigkeitsorientierte Materialflusskostenrechnung. Abwendung in

Lieferketten, der Abfallwirtschaft und Integration externer Effekte. Wiesbaden: Springer
Fachmedien, 2016.
Schuetz, T. (2016): Fahrzeugaerodynamik: Basiswissen für das Studium. Wiesbaden:
Springer Vieweg, 2016.
Scope, C.; Ilg, P.; Muench, S.; Guenther, E. (2016): Uncertainty in life cycle costing for
long-range infrastructure. part II: guidance and suitability of applied methods to address
uncertainty. In: International Journal of Life Cycle Assessment 21 (2016), p. 1170–1184.
Simpson, T. W. (2004): Product platform design and customization: Status and promise. In:
Artificial Intelligence for Engineering Design, Analysis and Manufacturing 18 (2004),
p. 3–20.
Smeets, E.; Weterings, R. (1999): Environmental Indicators: Typology and Overview. In:
European Environment Agency Technical report (1999), No 25, p. 1–19.
Sonego, M.; Echeveste, M. E.; Debarba, H. G. (2018): The role of modularity in sustainable
design: A systematic review. In: Journal of Cleaner Production 176 (2018), p. 196–209.
Sosa, M.; Eppinger, S.; Rowles, C. (2007): A Network Approach to Define Modularity of
Components in Complex Products. In: American Society of Mechanical Engineers 129
(2007), No. 11, p. 1118–1129.
Sphera (2019): Ökobilanz Datenbanken, retrieved from http://www.gabi-software.com/deu
tsch/datenbanken/.
Steubing, B.; Mutel, C.; Suter, F.; Hellweg, S. (2016): Streamlining scenario analysis and
optimization of key choices in value chains using a modular LCA approach. In: Interna-
tional Journal of Life Cycle Assessment 21 (2016), p. 510–522.
Stranddorf, H.; Hoffmann, L.; Schmidt, A. (2005): Impact categories, normalization and
weighting in LCA. Danish Ministry of Environment, Environmental Protection Agency
Environmental News 98, 2005.
Suhl, L.; Mellouli, T. (2013): Optimierungssysteme. Modelle, Verfahren, Software, Anwen-
dungen. 3 Edition. Berlin, Heidelberg: Springer, 2013.
Transportpolicy (2020a): EU: Light-Duty: GHG Emissions, 2020, URL: https://www.transp
ortpolicy.net/standard/eu-light-duty-ghg-emissions/, last checked on 14.06.2020.
Transportpolicy (2020b): China: Light-duty: Fuel Consumption, 2020, URL: https://
www.transportpolicy.net/standard/china-light-duty-fuel-consumption/, last checked on
14.06.2020.
Transportpolicy (2020c): US: Light-duty: Fuel Economy and GHG, 2020, URL: https://
www.transportpolicy.net/standard/us-light-duty-fuel-economy-and-ghg/, last checked on
14.06.2020.
Transportpolicy (2020d): California: Light-duty: GHG, 2020, URL: https://www.transport
policy.net/standard/california-light-duty-ghg/, last checked on 14.06.2020.
Tseng, H.-E.; Chang, C.-C.; Li, J.-D. (2008): Modular design to support green life-cycle
engineering. In: Expert Systems with Applications 34 (2008), p. 2524–2537.
Umweltbundesamt (2015): Umwelttrends in Deutschland—Daten zur Umwelt 2015. BGZ
Druckzentrum GmbH, 2015.
Umweltbundesamt (2016): Weiterentwicklung und vertiefte Analyse der Umweltbilanz von
Elektrofahrzeugen, Dessau-Roßlau: ifeu—Institut für Energie- und Umweltforschung
Heidelberg GmbH, 2016.
212 Literature

Umweltbundesamt (2019a): Emissionen ausgewählter Treibhausgase in Deutschland nach

Kategorien in Tsd. t Kohlendioxid-Äquivalenten, URL: https://www.umweltbundes
amt.de/sites/default/files/medien/384/bilder/dateien/8_tab_thg-emi-kat_2020.pdf, last
checked on 03.03.2020.
Umweltbundesamt (2019b): Produktverantwortung in der Abfallwirtschaft: Altfahrzeuge,
URL: https://www.umweltbundesamt.de/themen/abfall-ressourcen/produktverantwo
rtung-in-der-abfallwirtschaft/altfahrzeuge, last checked on 05.02.2019b.
Ulmschneider, M. (2004): Life Cycle Costing (LCC) und Life Cycle Assessment (LCA)—eine
Übersicht bestehender Konzepte und deren Anwendung am Beispiel von Abwasserpump-
stationen. Technische Universität Dresden, Dissertation, 2004.
Ulrich, K.: The role of product architecture in the manufacturing firm. In: Research Policy
24 (1995), No. 3, p. 419–440.
Umeda, Y.; Fukushige, S.; Tonoike, K.; Kondoh, S. (2008): Product modularity for life cycle
design. In: CIRP Annals—Manufacturing Technology 57 (2008), No. 1, p. 13–16.
UNEP (United Nations Environment Programme) (1996): Life Cycle Assessment: What it is
and how to do it. Paris: United Nations Publication, 1996.
UNEP IRP (United Nations Environment Programme International Resource Panel)(2017):
Assessing global resource use: A systems approach to resource efficiency and pollu-
tion reduction. In: Bringezu, S., Ramaswami, A., Schandl, H., O’Brien, M., Pelton, R.,
Acquatella, J., Ayuk, E., Chiu, A., Flanegin, R., Fry, J., Giljum, S., Hashimoto, S., Hell-
weg, S., Hosking, K., Hu, Y., Lenzen, M., Lieber, M., Lutter, S., Miatto, A., Singh
Nagpure, A., Obersteiner, M., van Oers, L., Pfister, S., Pichler, P., Russell, A., Spini, L.,
Tanikawa, H., van der Voet, E., Weisz, H., West, J., Wiijkman, A., Zhu, B., Zivy, R. A
Report of the International Resource Panel. United Nations Environment Programme.
Nairobi, United Nations Publication, 2017.
United Nations (1992): The Rio Declaration on Environment and Development, 1992, URL:
https://www.un.org/en/development/desa/population/migration/generalassembly/docs/
globalcompact/A_CONF.151_26_Vol.I_Declaration.pdf, last checked on 12.10.2018.
United Nations (2015a): Paris Agreement, 2015, URL: https://unfccc.int/files/essential_
background/convention/application/pdf/english_paris_agreement.pdf, last checked on
07.06.2020.
United Nations (2015b): Transforming our World. The 2030 Agenda for Sustainable Devel-
opment, 2015, URL: https://sustainabledevelopment.un.org/content/documents/212
52030%20Agenda%20for%20Sustainable%20Development%20web.pdf, last checked
on 07.06.2020.
VDA (Verband der Automobilindustrie e.V.) (2014): Unsere Werke—Nachhaltige Automo-
bilproduktion in Deutschland. Brandenburgische Universitätsdruckerei und Verlagsge-
sellschaft Potsdam mbh, 2014.
VDI (Verein Deutscher Ingenieure) (2005): VDI 2884 Beschaffung, Betrieb und Instandhal-
tung von Produktionsmitteln unter Anwendung von Life Cycle Costing. VDI-Gesellschaft
Produktionstechnik, 2005.
Vetter, P. (2017): Leichtbau fällt nicht schwer ins Gewicht. In: Welt am Sonntag (2017), No.
49, p. 36.
Volkswagen AG (2017): Transformation gestalten—Nachhaltigkeitsbericht 2017, URL:
https://www.volkswagenag.com/presence/nachhaltigkeit/documents/sustainability-rep
ort/2017/Nichtfinanzieller_Bericht_2017_d.pdf, last checked on 06.06.2020.
Literature 213

Volkswagen AG (2019a): Volkswagen Car Configurator, VW Golf Model, URL: https://

www.volkswagen.de/app/konfigurator/vw-de/de/der-golf/30315?page=trim, last checked
on 11.02.2019.
Volkswagen AG (2019b): Volkswagen Newsroom—Klimabilanz von E-Fahrzeugen & Life
Cycle Engineering, URL: https://uploads.volkswagen-newsroom.com/system/produc
tion/uploaded_files/14448/file/da01b16ac9b580a3c8bc190ea2af27db4e0d4546/Klimab
ilanz_von_E-Fahrzeugen_Life_Cycle_Engineering.pdf?1556110703, last checked on
24.07.2020.
Volkswagen AG (2020): WLTP Information, URL: https://www.volkswagenag.com/en/
group/fleet-customer/WLTP.html, last checked on 14.06.2020.
Wang, Q.; Tang, D.; Yin, L.; Yang, J. (2016): A Method for Green Modular Design Consid-
ering Product Platform Planning Strategy. In: Procedia CIRP 56 (2016), p. 40–45.
Wansart, J. (2012): Analyse von Strategien der Automobilindustrie zur Reduktion von CO2-
Flottenemissionen und zur Markteinführung alternativer Antriebe. Ein systemdynamis-
cher Ansatz am Beispiel der kalifornischen Gesetzgebung. Wiesbaden: Springer-Gabler,
2012.
WBCSD (World Business Council on Sustainable Development) (1995): Eco-Efficient
Leadership for Improved Economic and Environmental Performance. 1995, URL: http://
wbcsdservers.org/wbcsdpublications/cd_files/datas/wbcsd/business_role/pdf/EELeaders
hipForImprovedEconomic&EnviPerformance.pdf, last checked on 06.06.2020.
WCED (World Commission on Environment and Development) (1987): Report of the World
Commission on Environment and Development: Our Common Future. Oxford: Oxford
University Press, 1987.
Wedler, D.; Vietor, T. (2019): Potentials of modular autonomous vehicles for variable scenar-
ios of public transport. In: Bargende, Michael; Reuss, Hans-Christian; Wagner, Andreas;
Wiedemann, Jochen (eds) 19. Internationales Stuttgarter Symposium. Automobil- und
Motorentechnik. Wiesbaden: Springer Fachmedien, 2019, p. 589–602.
Weymar, E.; Finkbeiner, M. (2016): Statistical analysis of empirical lifetime mileage data for
automotive LCA. In: International Journal of Life Cycle Assessment 21 (2016), p. 215–
223.
Wilhelm, B. (1997): Platform and Modular Concepts at Volkswagen—Their Effects on the
Assembly Process. In: Shimokawa, Koichi; Jürgens, Ulrich; Fujimoto, Takahiro (eds)
Transforming Automobile Assembly. Berlin, Heidelberg: Springer, 1997, p. 146–156.
Yang, Q.; Yu, S.; Jiang, D. (2014): A modular method of developing an eco-product family
considering the reusability and recyclability of customer products. In: Journal of Cleaner
Production 64 (2014), p. 254–265.
Zimmermann, H.-J. (2008): Operations Research Methoden und Modelle. Für Wirtschaftsin-
genieure, Betriebswirte, Informatiker. 2nd Edition. Wiesbaden: Vieweg, 2008.