Journal of Manufacturing Systems 33 (2014) 51–64

Contents lists available at ScienceDirect

Journal of Manufacturing Systems

journal homepage: www.elsevier.com/locate/jmansys

Technical paper

A model for assessing the layout structural complexity of

manufacturing systems
H. ElMaraghy ∗ , T. AlGeddawy, S.N. Samy, V. Espinoza
Department of Industrial and Manufacturing Systems Engineering, University of Windsor, Windsor, ON, Canada

a r t i c l e i n f o a b s t r a c t

Article history: The layout of a manufacturing facility/system not only shapes its material flow pattern and influence
Received 11 October 2012 transportation and operation cost, but also affects logistics and parts/machine assignment decisions. The
Received in revised form 15 May 2013 layout of manufacturing systems determines its structural complexity by virtue of its design configuration
Accepted 28 May 2013
characteristics. This paper introduces a new model and indices for assessing the structural complexity
Available online 9 July 2013
of manufacturing systems layout in the physical domain. Six complexity indices, based on the physical
structural characteristics of the layout, have been introduced and formulated. They are layout density,
Keywords:
path, cycle, decision points, redundancy distribution and magnitude indices. An overall Layout Complex-
Layout
Complexity
ity Index (LCI) which combines all indices is developed using a novel method based on radar plots which
Manufacturing systems is insensitive to the order of plotting the individual indices. The use of the developed LCI is demon-
Graph theory strated using six typical types of manufacturing systems layouts and relevant guidelines are presented.
The developed model and complexity indices help design system layouts for least complexity and com-
pare layout alternatives that meet the specifications, at early design stages. It supports making trade-off
decisions regarding manufacturing systems flexibility and complexity and their associated costs.
© 2013 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction operation and performance of manufacturing systems. A good

layout contributes to the overall efficiency of operations and
Manufacturing companies often operate in a dynamic environ- can reduce by up to 50% the total operating expenses [2].
ment driven by changes in market conditions, customer demands, However, despite the attention given by researchers to measur-
product design and processing technology. Complexity and uncer- ing structural complexity in many fields [3–5], systems layout
tainty limit the effectiveness of conventional production control complexity has not been given much consideration. Manu-
and scheduling approaches. The scope of complexity is categorized facturers need to introduce new products regularly and with
into: part, product, system, and system of systems. Many disci- minimum disruption of operations. Managers often cope with
plines, such as mathematics, statistical physics, biology, medicine the inefficiencies of existing layouts with limited and local-
and social sciences as well as computer science and engineering ized fixes rather than undergo expensive and time consuming
face the challenge of quantitatively measuring complexity of a sys- layout redesign. Emerging trends in industry which influence
tem and defining its limits. The sources of complexity may include: systems layout include: delayed product differentiation, scalable
(i) size, (ii) coupling, (iii) variety, and (iv) multi-disciplinarity. Com- machines, movable machines, and distributed and modular layouts
plexity may be static (structural) or dynamic (operational). Static [6].
complexity is time-independent and is inherent in the structure The objective of this research is to introduce metrics for
of an engineered system. Dynamic complexity is time-dependent assessing the structural complexity of manufacturing systems lay-
and is concerned with the operational behavior of the system out by defining its characteristics and flow patterns that contribute
[1]. to the complexity of decisions made during system operation
Manufacturing systems layout generation and evaluation is related to its layout.
challenging and time consuming due to its multi-objective nature A manufacturing system layout is defined by the arrangement
and requires extensive data collection, analysis and synthe- of machines, areas or departments, their locations and connec-
sis. Systems layout has been an active research area for many tions between them [7]. A manufacturing system configuration
decades. Layout complexity has a significant impact on the is the set of constituent components, such as machines, worksta-
tions, transporters and controllers, etc., and the number and type
of each module that make up a manufacturing system and their
∗ Corresponding author. relationships which define the flow of work pieces between them
E-mail addresses: [email protected], [email protected] (H. ElMaraghy). [8].

0278-6125/$ – see front matter © 2013 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jmsy.2013.05.012
52 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

Fig. 1. Manufacturing system Structural Classification Code (SCC) developed by ElMaraghy [11].

A new methodology is proposed which develops a graphical element. However, neither the relative importance of the individ-
representation of the physical manufacturing system layout and ual metrics was discussed nor were they combined into a single
produces measurable complexity indices, based on the number, system complexity index. ElMaraghy [11] developed a manufac-
locations and connections of decision points within the system turing system Structural Classification and Coding system (SCC)
layout. These indices are used to evaluate the structural complex- that captures its structural complexity in the physical domain
ity of layout alternatives and identify potential structural layout which is inherent in various types of equipment found in manu-
problems at an early design stage. facturing systems as well as those used for storage and material
handling (Fig. 1). This chain type structural classification code has
2. Complexity in manufacturing systems been extended to include assembly specific structural features of
equipment used in assembly systems, such as dedicated assembly
Measures for static (structural) [9] or dynamic (operational) machines, parts grippers, feeders, orienting devices and handling
manufacturing systems complexity have been proposed in liter- equipment [12]. The SCC code uses attributes of the system machin-
ature [3]. Frizelle [9] defined static complexity as the expected ery and their physical structure and their operation and control
amount of information necessary to describe the state of a manu- characteristics to assess the system overall structural complexity
facturing system and is based on the probability of resources being of the system configuration. It also proposed means for measur-
in a certain state. Dynamic complexity was defined as the expected ing the layout complexity. The SCC was applied to many industrial
amount of information necessary to describe the state of system examples including comparing the layout complexity of three sug-
deviation from schedule due to uncertainty [10,3] by measuring the gested systems for machining the same surfaces of an automobile
difference between the actual and scheduled system performance. engine block [13].
Kim [4] proposed a heuristic approach to quantify the struc- Gabriel [5] investigated the internal static manufacturing com-
tural and operational system complexity based a series of system plexity, based on product line complexity, product structure and
complexity metrics which measure the relationships between sys- process complexity components. However, this complexity mea-
tem components, number of elements and the complexity of each sure did not consider layout, arguing that it was difficult to assess
 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64 53

Fig. 2. Illustration of different joints and interfaces in MHSs.

layout complexity due to lack of any evident quantifiable elements. the increase in the number of route options at any layout juncture
A system configuration can become complex, due to the number of and with the presence of more than one type of transporters. The
machine arrangements and connections. Koren et al. [14] studied interfaces at these junctures are, therefore, referred to as the “deci-
and represented systems arranged in serial, parallel, and hybrid sion points” in this paper. Increased number of decision points and
patterns and analyzed the effect of their layout on system perfor- branching at these points in order to implement the required mate-
mance in terms of reliability, productivity, quality and scalability. rial flow patterns and achieve dynamic routing and enable systems
Hu et al. [15] measured the manufacturing complexity induced plug and play reconfiguration. This increases the number and intri-
by product variety in assembly systems and supply chains, by cacy of control decisions to be made at each node and consequently
developing models to characterize the propagation of complexity the system layout complexity. New designs for improving control
in multi-state mixed-model assembly systems and multi-echelon decisions made at intersections of material handling equipment
assembly supply chains. A multi-echelon system has a tree- have been developed which further emphasize the importance of
structure with multiple feeds at each station from predecessor taking the number of nodes in a layout and associated flow patterns
stages and a single feed to successor stages. A complexity model at these nodes into consideration in assessing its structural com-
based on product variety including station level complexity, sys- plexity. Fig. 3 illustrates an example of grouping small conveyor
tem level complexity and supply chain complexity was defined and modules to solve a routing problem within a factory transportation
used to configure assembly systems and supply chains to ensure system. The state of individual MHS module contains information
robust performance by mitigating complexity. In axiomatic design, from sensors located in each [18]. The number and nature of deci-
information and complexity are defined only relative to what we sions made at each joint grows significantly when several modes of
are trying to achieve and/or want to know. Suh [16] was the first transportation are used and interfaced within the same layout. This
to define complexity in the functional domain as a measure of adds to the complexity and cost of controlling flow at the juncture
uncertainty in achieving the specified Functional Requirements points (joints).
(FR). Complexity would be related to information content, which
is defined as a logarithmic function of the probability of achieving
the functional requirement. Therefore, complexity is proportional
to the amount information required to achieve the specified func-
tional requirement.
Manufacturing systems have numerous components and sub-
systems with many complex interactions and relationships
between them which increase its complexity and its information
content. The layout of the manufacturing system not only affects
the performance (i.e., productivity, throughput and quality) [17],
but also the structural complexity of the system which is inherent in
the number and type of connections between machines and mate-
rial handling systems, their cost and control logic and effort [5,1].
The operation and control of material handling systems (MHS)
illustrates, and is closely associated with, the layout-induced
structural system complexity. Most of MHSs are mechanically pre-
defined and, therefore, functionalities are locally built into their
construction and control features (Fig. 2). The MHS requirements
become more complex from the mechanical point of view con-
cerning the material to be conveyed, its functionality as well as
the system control [18]. To increase the flexibility of MHS, many
similar MHS modules are interfaced at several points to create
multi-directional material flow functionality and ability. However, Fig. 3. Grouping of individual modules in a conveyor matrix to transport large units.
the required sensing and control logic grow in complexity with Adapted from [18].
54 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

Fig. 4. An IDEF0 for the Layout Complexity Assessment (LCA) Model [19].

The proposed layout complexity model and indices assess the 1. System Layout Analysis: The layout is analyzed and a diagram
structural complexity of the manufacturing systems layout and is generated with nodes and arrows representing the decision
incorporate the information content of the system, represented by making points, and the connections in the system respectively.
the characteristics of its layout. They measure the effect of complex- 2. Adjacency Matrix Creation: A layout adjacency matrix is created
ity on the time, effort and cost of decisions made on the shop floor to capture the represented nodes and relationships between
to control material flow in a dynamic and responsive manner. The them in the diagram.
characteristics of the layout are expressed by various complexity 3. Complexity Indices Generation: Layout complexity indices (i.e.
indices as explained in the following sections. density, paths, cycles, decision points, redundancy distribu-
tion and magnitude) were developed to assess various aspects
3. Layout complexity assessment model of layout complexity. An overall layout structural complexity
index (LCI) has also been developed to combine the individ-
The proposed Layout Complexity Assessment (LCA) model con- ual indices. The density and number of decision point’s indices
sists of three steps as shown by an IDEF0 representation as shown in a layout correspond to the number and location of joint
in Fig. 4: interfaces between its modules, while paths, cycles, redundancy

Fig. 5. The conversion of manufacturing system layout to a layout matrix representation.

 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64 55

Fig. 6. Complexity indices of manufacturing systems layout (Adapted from [21]).

distribution and magnitude indices capture the pattern of mate- 3.3. Layout complexity indices generation
rial flow between nodes.
Six complexity indices related to nodes and arrows from the
graph representation are generated in this step of the LCA model.
3.1. System layout analysis A method of quantification is proposed for each index individu-
ally. The complexity indices are defined and then integrated into
The purpose of this step of the LCA model is to identify where an overall layout complexity index. The complexity indices are:
decisions are made based on the connections between differ- Density, Path, Cycle, Decision Points, Redundancy Distribution, and
ent manufacturing system components, areas and departments. Magnitude indices as shown in Fig. 6. They measure informa-
Manufacturing systems layouts are graphically represented by tion content which increases the difficulty of making decisions
nodes and arrows (Fig. 5). Nodes represent points where deci- regarding the flow of material within the system layout. The dif-
sions regarding material flow direction and destinations are made. ferent types of complexity indices are based on graph theory using
For instance, a choice of one of several alternate routes for the characteristics of graphs such as connectivity, paths and cycles [20].
next process is a decision to be made upon completion of pro- All indices are normalized between 0 and 1 representing the
cessing at a workstation. The arrows represent the connection minimum and theoretical information content and complexity
and direction of material flow (forward or backward) that exist respectively. The six complexity indices used in the LCA model are
between nodes. Two types of extra nodes are added to the graph defined and mathematically introduced in Sections 3.3.1 to 3.3.6.
for any layout being modeled (Fig. 5, part b): (1) Systems Input Simple layout examples are used to illustrate the application of the
nodes symbolizing the material entry points, and (2) System Out- developed complexity indices.
put nodes symbolizing material exit points from the system. The
inclusion of input and output nodes in the diagram helps identify 3.3.1. Density index
the total individual paths starting from one of the input nodes, The density of a layout graph is defined as the ratio of the number
through specific nodes, until reaching one of the output nodes. of graph nodes to the theoretical maximum number of those nodes,
Several input and output nodes may exist within one system lay- which is obtained by connecting all pairs of existing nodes. The
out. layout density index is calculated by Eq. (1).
k
D= (1)
n(n − 1)
3.2. Layout adjacency matrix creation
where k is the actual number of connections and n is the number
An Adjacency Matrix is generated using the nodes from the of nodes. A high density index is indicative of a more complex sys-
graph representation diagram shown at the header row and col- tem because of the relative increase in the number of connections
umn. The values in the matrix cells correspond to the arrows on compared to the number of nodes in the system layout. Examples
the graph representation. If two nodes are connected, which means of density index evaluation are shown in Fig. 7a and b.
that material flow exists between the represented workstations,
then the value is “1”; otherwise it is “0”. If many parallel connec- 3.3.2. Path index
tions exist between two nodes, the number of connections is stated A path in a layout graph is defined by the sequence of nodes
instead of “1”, such as the A-B connection in Fig. 5, part c. The beginning from input node following the arrows to the output node,
sequence of nodes placement in the matrix starts with the input following at least one arrow [22]. A path cannot cross the same
nodes and ends with the output nodes, and passes through other node twice. The number of paths is related to the structural com-
nodes in-between. Consequently, a square matrix of size n x n is plexity since it increases the number of alternate paths (routes)
created, where n is the number of nodes. The adjacency matrix is and decisions needed to control the flow within the system. Path
used to mathematically generate the various complexity indices in index compares the actual number of paths, which can be calcu-
the LCA model. lated using the paths finder algorithm, to the minimum theoretical
56 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

Fig. 7. Examples of layout complexity indices evaluation.

 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64 57

1. Obtain n: number of nodes of the layout A: Adjacency matrix of

the nodes
2. Initiate C ← 0: number of cycles
3. l ← 2
4. D ← Al .*I: evaluation matrix for l node cycles
.*: element-wise multiplication operator
I: identity matrix
5. x ← 0
6. If dij > 0 v i and j, then x ← x + 1
7. If x = 1, then C ← C + 1: a cycle of l nodes has been found
8. Repeat from step 4 until l = n: max number of nodes a cycle
9. Declare C: total number of cycles of all possible lengths

3.3.4. Decision points index

The decision points index represents the cumulative complexity
Fig. 8. Triangles formed in a radar chart representation. of decision making which increases with the number of nodes in
a path (i.e. the sum of all nodes between input and output nodes)
number of paths, which depends only on the number of input and [24]. The number of nodes per path is related to layout structural
output nodes. Path index is generated by Eq. (2). Examples of path complexity because it increases the number of decisions to be made
index evaluation are shown in Fig. 7c and d. along the path as well as the potential of system errors. The decision
p point index is calculated by Eq. (5).
P =1− (2)
N SP
DS = 1 − (5)
where p is the minimum theoretical number of paths = number of LP
input nodes × number of output nodes, N is the number of existing where SP is the number of nodes on the shortest path in the layout
paths calculated from the paths finder algorithm. graph, representing the theoretical minimum number of decision
Paths finder Algorithm: points in one sequence, while LP is the number of nodes on the
longest path, representing the actual number of decision points in
1. Obtain n: number of nodes of the layout one sequence. A high decision point index represents higher lay-
A: Adjacency matrix of the nodes out complexity due to the increase of number of decisions made
2. Initiate N ← 0: number of paths per path. SP and LP can be obtained during the execution of the
3. l ← 2 path finder algorithm by recording the least and the most number
4. B ← Al : evaluation matrix for l-1 node paths of nodes at which a path exists. Examples of index evaluation are
5. If binputs outputs = 1 V input nodes: relationship elements between shown in Fig. 7g and h.
input and output nodes
Then N = N + 1: a path of length l–1 has been found 3.3.5. Redundancy distribution index
6. l ← l + 1 Distribution of redundancy refers to the existence of arrows
7. repeat from step 4 until l = n: max number of nodes in a path between adjacent nodes, regardless of the number of the redun-
8. Declare N: total number of paths of all possible lengths dant arrows. Redundancy increases the information content in the
system layout, since decisions must be made to choose one of the
alternatives at those locations, hence, increasing the complexity.
3.3.3. Cycle index The redundancy distribution index compares the actual number
By definition, a cycle is a loop of nodes that starts and ends at of locations where there exist redundancy and the theoretical
the same node [23]. The length of a cycle is the number of the maximum number of redundancy locations, which is equal to the
nodes within its loop. A cycle adds structural complexity to the number of adjacencies between nodes. Redundancy Distribution
system, since the flow does not follow a linear sequence and recur- Index is calculated by Eq. (6).
rent flows increase the difficulty to follow material flow across
r
system layout. The complexity index of the cyclic effect of node RD = (6)
a
loops compares the actual number of cycles in the layout graph rep-
resentation and compares it to the theoretical maximum number where r is the number of locations where there are redundant
of cycles. A large number of cycles indicates high recurrent flows, arrows, and a is the theoretical maximum number of locations in
hence, higher layout complexity. Cycle index is presented by Eqs. the graph representation where redundancy may exist. Examples
(3) and (4). Examples of cycle index evaluation are shown in Fig. 7e of redundancy distribution index evaluation are shown in Fig. 7i
and f. and j.
C
CL = (3) 3.4. Redundancy magnitude index
MC
where C is the actual number of cycles calculated from the cycles The redundancy magnitude index accounts for the number of
finder algorithm redundant parallel arrows in the system layout. This is related to

n the layout structural complexity as the number of redundant par-
MC = Cin (4) allel arrows increases the information content. The redundancy
i=2 magnitude index is calculated by Eq. (7).
where MC is the theoretical maximum number of cycles. Cin is the pr w−a
RM = = (7)
combination of n nodes starting with two nodes, since at least a w w
pair of nodes is needed to have a cycle where pr is the total number of redundant paralell arrows; w is
Cycle Finder Algorithm: the total number of forward arrows, representing the theoretical
58 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

Fig. 9. Engine piston assembly layout (Adapted from [25]).

maximum number of redundant arrows, while a is the number of 4.1. The LCI formulation
adjacent nodes. Examples of index evaluation are shown in Fig. 7k
and l. A new formulation for aggregating the six layout complexity
indices which avoids the sequence dependence drawback of Eq.
(8) has been developed. It is a general formula for calculating the
4. Overall layout complexity assessment
average area of all possible areas formed by all possible sequences
of plotting the complexity indices on a radar plot. Each two indices
The system layout complexity assessment (LCA) index is the
form a triangle on the radar chart polygon as illustrated in Fig. 8.
aggregation of the individual complexity indices into an overall
The area of the shaded triangle in Fig. 8 for indices X1 and X2 is
complexity index. This is needed to obtain a single index that rep-
calculated by Eq. (9).
resents the overall system layout complexity which can be used
to compare various layout design alternatives. A radar plot, where sin 
A= C1 C2 (9)
indices are uniformly distributed and plotted is used to combine 2
all indices as illustrated in Fig. 8. Despite the assumption of equal where  is the angle between the two indices.
weights, the indices are treated independently and normalized by Consequently, the general formula to expresses the area fromed
its maximum value on each axis on the radar plot. For the purpose by any pair of indices Ci and Cj is expressed by Eq. (10):
of comparing system alternatives, the assumption of equal weights
is acceptable. However, is possible for system designers to decide sin 
A= Ci Cj (10)
the relative importance of each individual complexity index. A for- 2
mula to calculate the shaded area in a radar plot was developed by Let the first position of that triangle be fixed, while the other indices
Samy [25]. The shaded area is calculated by Eq. (8). take values from Ci+1 to Cn . Then there exist (n − 1) triangles for each
  index Ci , where n is the number of indices. The average area of those
1 
i=t−1  360  triangles can be expressed by Eq. (11).
a= (Ct ∗ C1 ) + (Ci ∗ Ci+1 ) sin (8) ⎛ ⎞
2 t
i=1
⎜ ⎟
⎜n ⎟
where a is the shaded radar area, Ci is the individual complexity sin  ⎜ ⎟
Ai = C ⎜ Cj ⎟ (11)
index value on the radial axis of index i and t is total number of com- 2(n − 1) i ⎜ ⎟
plexity indices. However, this formula is sensitive to the sequences ⎝j = 1 ⎠
of the individual indices on the radar plot [21]. The radar plot is j=
/ i
a good tool to compare system alternatives visually by inspecting
the shaded areas on the radar plot. But, in many cases it is not easy The i index in Eq. (11) is summed from 1 to n to account for all
to compare areas visually especially for multivariate data. The area possible indices combinations. The average area of all triangles in
is strongly influenced by the order in which the indicators are dis- the radar chart can then be expressed by Eq. (12).
played and hence affects the comparison and it is a deficit of the ⎛ ⎞
radar plot.
⎜ ⎟
⎜ ⎟
sin   ⎜
n n
⎟
Aavg = Ci ⎜ Cj ⎟ (12)
2(n − 1) ⎜ ⎟
i=1 ⎝j = 1 ⎠
j=
/ i

Eq. (12) can be further simplified as follows:

⎛ ⎞
sin    n n
A= Ci ⎝ Cj − Ci ⎠ (13)
2(n − 1)
i=1 j=1
Fig. 10. Graphical representation of the engine piston assembly plant layout.
 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64 59

Fig. 11. Automotive engine belt tensioners DPD assembly system layout (Adapted from [26]).

Hence: variant since it aims at studying the effect of increasing product

⎛ ⎞ variety on system layout complexity.
sin  
n

n 
n
A= ⎝ Ci Cj ⎠ − Ci Ci (14)
2(n − 1) 5.1. Automobile engine piston assembly plant layout
i=1 j=1 i=1

This plant is dedicated to the assembly of two types of automo-

Consequently:
bile engine pistons [25]. The assembly process starts with a gantry
n 2 robot that picks up the piston head by suction. Parts, including the
 
n

LCI = Ci − Ci2 (15) connecting rod, the piston pin, and the snap rings, are moved to
the indexing table which feeds the press. Then, one press inserts
i=1 i=1
the pins into the piston head. The next press inserts the snap rings
The term sin/2(n − 1) has been removed since it is constant for into the piston head. This is followed by inspection to check the
the same number of indices. The Overall system Layout Complexity presence of the snap rings, otherwise the subassembly is returned
Index (LCI) is insensitive to the sequence of the individual complex- for re-work. The piston sub-assembly is placed next on an overhead
ity indices. conveyor. The piston rings are located in the cylinderical magazines

5. Applications and case studies

The developed Layout Complexity Index (LCI) is applied sys-

tematically to four typical manufacturing system layouts to assess
their structural complexity. Each layout represents a class of manu-
facturing systems layout. They include: (1) The automobile engine
piston assembly plant is a dedicated manufacturing system, with
fully automated material handling system and dedicated equip-
ment, (2) A Delayed Product Differentiation layout that postpones
the configuration of engine belt tensioner product variants, (3) A
process layout with dedicated process departments which produce
automotive brake systems, and (4) A cellular manufacturing layout
with more process flexibility for the same automotive brake sys-
tems. All of these layouts can be mathematically represented by
five nodes, hence providing a unified basis for comparison, except
for the second one, which has six nodes, one for each product Fig. 12. Graphical representation for the hybrid DPD assembly system layout.
60 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

Fig. 13. Brake Systems – Process (BSP) layout (Adapted from [21]).

and are inserted into the piston head manually. A nut runner dis- Individual complexity indices are calculated as:
assembles the connecting rod cap. The bearings are fed into a pick D = 8/(5(5 − 1)) = 0.4, P = 1 − (2/2) = 0, C = 4/26 = 0.15,
and place device by a conveyor. Later, the bearings are placed in DS = 1 − (3/3) = 0, RD = 0/4 = 0, RM = 0/8 = 0
the position to be pressed into the subassembly. The robot picks up
the finished assembly and places it on a pallet on a belt conveyor, n 2
 
n
as illustrated in Fig. 9. LCI = Ci − Ci2 = 0.31 − 0.18 = 0.12
The diagram representation of the engine piston assembly sys-
i=1 i=1
tem layout is shown in Fig. 10.
The corresponding piston assembly system layout adjacency
matrix is:
5.2. Delayed product differntiation (DPD) engine belt tensioner
assembly system layout

In A B C D E O1 O2 This layout demonstrates an example of Delayed Product Dif-

⎛ ⎞ ferentiation (DPD) of automotive engine belt tensioners [26]. This
In 0 1 0 0 0 0 0 0
type of flow pattern is often used in mass-customization assem-
A⎜0 0 1 1 0 0 0 0⎟
⎜ ⎟ bly systems [6]. The shown DPD layout is a hybrid serial-parallel
B ⎜0 0 0 0 1 0 0 0⎟
⎜ ⎟ layout consisting of two processing stages and one differentiation
C ⎜0 1 0 0 0 1 0 0⎟ point for five belt tensioner variants. In the first stage, process plat-
AM = ⎜ ⎟
D ⎜⎜0 0 1 0 0 0 1 0⎟
⎟ form A in Fig. 11, un-differentiated sub-assemblies are constructed.
E ⎜0 1⎟ Different variants are customized beyond the point of product dif-
⎜ 0 0 1 0 0 0 ⎟
O1 ⎝ 0 0⎠
ferentiation, based on actual demand. Each differentiation process
0 0 0 0 0 0
branch (B, C, D, E and F) has an inspection station where the product
O2 0 0 0 0 0 0 0 0 is either accepted and removed or returned to the point of product
 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64 61

The adjacency matrix is created according to the relationship

between adjacent nodes in Fig. 14.

In A B C D E O1 O2
⎛ ⎞
In 0 1 0 0 0 0 0 0
⎜0
A 0 7 0 0 0 0 0⎟
⎜ ⎟
⎜0
B 0 0 14 0 0 0 0⎟
⎜ ⎟
C ⎜0 0 0 0 1 1 0 0⎟
AM = ⎜ ⎟
Fig. 14. BSP layout graphical representation.
D ⎜⎜0 0 0 1 0 0 1 0⎟
⎟
E ⎜0 1⎟
⎜ 0 0 1 0 0 0 ⎟
O1 ⎝ 0 0 0 0 0 0 0 0⎠
O2 0 0 0 0 0 0 0 0
differentiation for re-work. Figs. 11 and 12 show the layout of the
system and its graph representation respectively. The adjacency matrix is used as the input for path and
The corresponding system layout adjacency matrix is: cycle finder algorithms to define the characteristics related
to the complexity indices. Individual complexity indices are
calculated as D = 6/(5(5 − 1)) = 0.3, P = 1 − (2/2) = 0, C = 2/26 = 0.08,
In A B C D E F O1 O2 O3 O4 O5 DS = 1 − (4/4) = 0, RD = 2/4 = 0.5, RM = 19/23 = 0.83.
⎛0 1 ⎞
In 0 0 0 0 0 0 0 0 0 0 n 2
A ⎜0 0 2 1 1 1 1 0 0 0 0 0⎟
 
n

⎜ ⎟ LCI = Ci − Ci2 = 2.90 − 1.03 = 1.87

B ⎜0 1 0 0 0 0 0 1 0 0 0 0⎟
⎜ ⎟ i=1 i=1
C ⎜0 1 0 0 0 0 0 0 1 0 0 0⎟
⎜ ⎟
D ⎜0 1 0 0 0 0 0 0 0 1 0 0⎟
⎜ ⎟ 5.4. Brake systems – cellular (BSC) layout
E ⎜0 1 0 0 0 0 0 0 0 0 1 0⎟
AM = ⎜ ⎟
F ⎜0 1 0 0 0 0 0 0 0 0 0 1⎟
⎜ ⎟ The original process layout plant in Section 5.3 has been mod-
O1 ⎜0 0 0⎟ ified to adopt a cellular layout with increased flexibility and
⎜ 0 0 0 0 0 0 0 0 0 ⎟
O2 ⎜0 0 0⎟ responsiveness. This modified plant of BSC layout has 4 manufac-
⎜ 0 0 0 0 0 0 0 0 0 ⎟
O3 ⎜0 0 0⎟
turing cells; 3 of them are for ABS products and one cell is used
⎜ 0 0 0 0 0 0 0 0 0 ⎟
O4 ⎝0 0 0 0 0 0 0 0 0 0 0 0⎠
for both ABS and ASR. Each cell includes dedicated machining and
deburring machines. The new assembly area has 3 lines: 2 of these
O5 0 0 0 0 0 0 0 0 0 0 0 0 are for ABS and one line is for both ABS and ASR. Parts that do
not meet specifications in assembly line inspection, are returned
Individual complexity indices are calculated as back to the relevant manufacturing cell. The Layout and graph rep-
D = 10/(6(6 − 1)) = 0.33, P = 1 − (5/5) = 0, C = 5/57 = 0.09, resentation of the modified layout are shown in Figs. 15 and 16
DS = 1 − (2/2) = 0, RD = 1/5 = 0.2, RM = 1/6 = 0.17. respectively.
The corresponding adjacency matrix for the BSC layout is:
n 2 In A B C D E O1 O2
 
n
⎛ ⎞
LCI = Ci − Ci2 = 0.62 − 0.19 = 0.43 In 0 1 0 0 0 0 0 0
⎜0
A 0 1 3 0 0 0 0⎟
i=1 i=1 ⎜ ⎟
⎜0
B 0 0 0 1 0 0 0⎟
⎜ ⎟
C ⎜0 0 0 0 1 2 0 0⎟
AM = ⎜ ⎟
D ⎜⎜0 0 1 1 0 0 1 0⎟
⎟
E ⎜0 1⎟
5.3. Brake systems – process (BSP) layout ⎜ 0 0 1 0 0 0 ⎟
O1 ⎝ 0 0 0 0 0 0 0 0⎠
The facility shown in Fig. 13 is a brake system assembly plant O2 0 0 0 0 0 0 0 0
for automotive applications. This example is intended to show the
effect of process layout application on layout complexity. This plant Individual complexity indices are calculated as D = 8/(5(5 − 1)) = 0.4,
of BSP layout produces two product variants: the anti-lock braking P = 1 − (2/3) = 0.33, C = 6/26 = 0.27, DS = 1 − (3/5) = 0, RD = 2/5 = 0,
system (ABS) and anti-lock braking system with traction control RM = 3/8 = 0.38.
(ASR). Parts are first moved by a conveyor belt to the washing and n 2
 
n
drying machines, then Automated Guided Vehicles (AGVs) are used LCI = Ci − Ci2 = 4.75 − 0.81 = 3.95
to transport the part containers to one of the 7 identical machining
i=1 i=1
cells. After machining, the parts are moved to one of 14 identical de-
burring machines by AGVs to remove burrs using a high pressure
water jets. The assembly tasks are performed in two separate U- 6. Results and discussion
shaped cells for ABS and ASR products respectively, at which there
is an inspection to return parts if not cleaned satisfactorily back to The generated adjacency matrix, in each system, is used as the
the de-burring area. Accepted finished parts are manually taken off input to the path and cycle finder algorithms as described in Sec-
the conveyor and sent to the shipping area. The physical system tion 3.3 to calculate the characteristics related to the individual
layout is analyzed and the decision points and material flow direc- complexity indices. All four system layouts include 5 nodes except
tions are identified. Subsequently, a diagram representation of the the DPD system that has 6 nodes. The layout complexity indices
layout is generated as shown in Fig. 14. of the four assessed layouts are then calculated and summarized
62 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

Fig. 15. Brake Systems – Cellular (BSC) assembly layout (Adapted from [21]).

in Table 1. The analyzed systems layouts and the corresponding more nodes compared with the engine assembly system due to
LCIs are presented in an ascending order based on their complexity number of product variants involved (total of 5 variants). This did
value. Results obtained from Formula (15) are shown. The values not affect the density of the layout, however, the layout scored high
of overall system layout complexity index are ranked from 1 to 4, in redundancy category due to the parallel connections from node
where “1” indicates the least complex manufacturing system layout A to node B (Fig. 12), resulting in an increased overall system layout
and “4” represents the most complex system layout. complexity (0.43).
It can be seen that the resulting complexity order for the The results also indicate an increased layout complexity for cel-
assessed systems is coinciding with the nature of their systems and lular manufacturing layout compared to the process layout. The
inherent decision making processes based on their layout charac- BSC system of the cellular layout has the highest layout complex-
teristics. In the engine piston assembly plant, dedicated automation ity index among all studied layouts. The system features increased
indicates less decision making due to the fixed and pre-defined routing flexibility due to the existence of increased number of alter-
routing. The graphical representation in Fig. 10 shows almost a nate paths (4). This demonstrates that increasing the flexibility by
linear machines arrangement and material flow. The engine pis- adding alternate routings also increases the overall layout complex-
ton assembly plant scored the least layout complexity (0.12), since ity. On the other hand, the process layout BSP system has only one
it does not have redundant connections. As a result redundancy path but with numerous redundant connections, scoring the high-
indices have zero values. The belt tensioners DPD system has two est values for the distribution redundancy index and the magnitude
redundancy index among all layouts (0.5 and 0.83 respectively) but
less overall complexity (1.87) than the BSC layout (3.95).
In all previous layouts, it can be seen that the value of Den-
sity index is almost constant (from 0.3 to 0.4), since the number of
existing nodes in most of the layouts is the same (5 nodes: A, B, C,
D, and E), and the number of non-redundant connections among
nodes did not change much from one layout to another. All the
assessed layouts scored 0 for the decision point index, since the
longest and shortest paths are the same, except the cellular manu-
facturing layout. The BSC layout has short paths (3 nodes) and long
path (5 nodes) due to the existence of alternate routing.
Fig. 16. Graphical representation for the BSC assembly system layout.
 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64 63

Table 1
Complexity indices for the four assessed assembly systems layouts.

Rank System name Layout type Density index Path index Cycle index Decision point Distribution Magnitude redundancy LCI
inde‘x redundancy index index

1 Engine Assembly Dedicated 0.40 0 0.15 0 0 0 0.12

2 DPD Postponement 0.33 0 0.09 0 0.20 0.17 0.43
3 BSP Process 0.30 0 0.08 0 0.50 0.83 1.87
4 BSC Cellular 0.40 0.33 0.27 0.4 0.40 0.38 3.95

7. Discussion and conclusions structural manufacturing systems complexity makes it feasible to

integrate it into the overall design process.
New manufacturing paradigms and technology trends call for
more modularization of products and manufacturing platforms,
Acknowledgements
delayed product differentiation for mass customization, and more
equipment to enhance production flexibility, scalability, reconfig-
The authors would like to acknowledge the support and funding
urability and changeability. This gives rise to new types systems
provided by the Natural Sciences and Engineering Research Council
layouts with higher degree of complexity due to the nature of their
(NSERC) of Canada and the Canadian Research Chair (CRC) program.
structure and increased decisions made at every node within the
layout.
A new model for assessing the structural complexity of a man- References
ufacturing system layout was developed. Six individual layout
[1] ElMaraghy W, ElMaraghy H, Tomiyama T, Monostori L. Complexity in engi-
complexity indices which capture characteristics of the layout, such
neering design and manufacturing. CIRP Annals – Manufacturing Technology
as connections, paths, cycles, decision points, and redundancies 2012;61:793–814.
were defined and formulated. These indices capture and represent [2] Hasan MA, Sarkis J, Shankar R. Agility and production flow layouts: an analytical
the information content and structural complexity inherent in the decision analysis. Computers and Industrial Engineering 2012;62(4):898–907.
[3] Calinescu A, Efstathiou J, Schirn J, Bermejo J. Applying and assessing two meth-
manufacturing system layout. They are useful at the early design ods for measuring complexity in manufacturing. Journal of the Operational
stage when designing systems layouts and selecting connections, Research Society 1998;49(7):723–33.
paths, cycles and redundancies in the processes as well as the type [4] Kim Y-S. A System Complexity Approach for the Integration of Product
Development and Production System Design. Master of Science Depart-
of material handling systems. ment of Mechanical Engineering, Massachussetts Institute of Technology;
The individual complexity indices are combined into an over- 1999.
all Layout Complexity Index LCI. Aggregation of individual indices [5] Gabriel AJ. The effect of internal static manufacturing complexity on man-
ufacturing performance. Industrial Management, Clemson University; 2007.
into an overall complexity index makes it easier to compare system Ph.D.
layout alternatives and make decisions to select the least complex [6] Benjaafar S, Heragu SS, Irani SA. Next generation factory layouts: research chal-
layout. The layout complexity index was determined based on a lenges and recent progress. Interfaces 2002;32(6):58–76.
[7] Spicer P, Koren Y, Shpitalni M, Yip-Hoi D. Design principles for machin-
new formulation using a radar chart. The developed layout com- ing system configurations. CIRP Annals – Manufacturing Technology
plexity indices aggregation formula is independent of the order of 2002;51(1):275–80.
plotting the individual complexity indices on the radar chart which [8] Youssef AMA, ElMaraghy HA. Assessment manufacturing systems reconfigura-
tion smoothness. International Journal of Advanced Manufacturing Technology
is a significant advantage.
2006;30:174–93.
Guidelines such as reducing number of cycles, density and deci- [9] Frizelle G, W.E. Measuring complexity as an aid to developing operational
sion points are recommended to reduce manufacturing systems strategy. International Journal of Operations & Production Management
layout complexity. Two of the presented layout examples showed 1995;15(5):26–39.
[10] Deshmukh AV, Talavage JJ, Barash MM. Complexity in manufacturing sys-
an association between increasing system flexibility by adding new tems, part 1: analysis of static complexity. IIE Transactions 1998;30(7):
components and the increase of its structural complexity. Designers 645–55.
should be aware of the consequences of those design and configu- [11] ElMaraghy HA. A complexity code for manufacturing systems. In: ASME Confer-
ence Proceedings. Ypsilanti, MI, United states: American Society of Mechanical
ration decisions on system complexity. Engineers; 2006.
Certain manufacturing systems layouts intended to reduce trav- [12] ElMaraghy H, Samy SN, Espinoza V. A classification code for assembly systems.
eled distances and minimize cycle time and cost, such as the star In: 3rd CIRP Conference on Assembly Technologies and Systems, CATS2010.
2010.
or hub and spoke layouts, create nodes with many branches which [13] ElMaraghy HA, Kuzgunkaya O, Urbanic RJ. Manufacturing systems config-
increase the complexity of decision making at these nodes. They uration complexity. CIRP Annals – Manufacturing Technology 2005;54(1):
should be compared with alternate layouts that could achieve 445–50.
[14] Koren Y, Hu SJ, Weber TW. Impact of manufacturing system configura-
the same end result with less ongoing operational complex- tion on performance. CIRP Annals – Manufacturing Technology 1998;47(1):
ity. 369–72.
The new and novel layout complexity index (LCI) and devel- [15] Hu S, Zhu X, Wang H, Koren Y. Product variety and manufacturing complexity in
assembly systems and supply chains. CIRP Annals – Manufacturing Technology
oped assessment approach represent a decision making support
2008;57(1):45–8.
tool to measure the structural complexity of manufacturing sys- [16] Suh NP. A theory of complexity: periodicity and the design axioms. Research
tems layouts in the physical domain, which in turn affects the in Engineering Design 1999;11(2):116–32.
complexity and cost of their day to day operation. The developed [17] Huang H. Facility layout using layout modules. Industrial and Systems Engi-
neering, Ohio State University; 2003. Ph.D.
complexity indices would also help design layouts for less complex- [18] Krühn T, Falkenberg S, Overmeyer L. Decentralized control for small-scaled
ity. They can support the trade-off decisions to be made between conveyor modules with cellular automata. In: Proceedings of the 2010 IEEE
layout flexibility and complexity which impacts the cost of acquir- International Conference on Automation and Logistics (ICAL). 2010.
[19] Espinoza V, ElMaraghy HA, AlGeddawy T, Samy SN. Assessing the Structural
ing and constructing the system layout as well as the cost of its Complexity of Manufacturing Systems Layout. In: 4th CIRP Conference on
operation. Layout complexity is one of many factors influencing Assembly Technology and Systems – CATS. 2012.
the design of manufacturing systems layouts and choice of material [20] Bondy JA, Murty USR. Graph theory. New York: Springer; 2008.
[21] Espinoza Vega VB. Structural Complexity of Manufacturing Systems Layout.
handling systems and should be part of a comprehensive optimiza- Canada: Department of Industrial and Manufacturing Systems Engineering,
tion and trade-off analysis. Having a single metric representing the University of Windsor; 2012. M.Sc. Thesis.
64 H. ElMaraghy et al. / Journal of Manufacturing Systems 33 (2014) 51–64

[22] McCabe TJ. A complexity measure. IEEE Transactions on Software Engineering [25] Samy SN. Complexity of Products and Their Assembly Systems. Department
1976;2(4):308–20. of Industrial and Manufacturing Systems Engineering: Windsor; 2011. PhD
[23] Badke-Schaub, Gehrlicher A. Patterns of decision in design: leaps, loops, cycles, Dissertation.
sequence and meta-processes. In: Proceedings of the 14th International Con- [26] AlGeddawy T, ElMaraghy H. Design of single assembly line for the delayed
ference on Engineering Design. 2003. differentiation of product variants. Flexible Services and Manufacturing Journal
[24] Newman ME. The structure and function of complex networks. SIAM 2011;22(3):163–82.
2003;45:167–256.