Applied Thermal Engineering 212 (2022) 118533

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Research Paper

Thermal response analysis and parameter prediction of additively

manufactured polymers
Navid Moslemi a, Behzad. Abdi b, *, Soheil Gohari c, Izman Sudin a, E. Atashpaz-Gargari d,
Norizah Redzuan a, Amran Ayob a, Colin Burvill c, Meini Su b, Farid Arya e
a
School of Mechanical Engineering, Universiti Teknologi Malaysia (UTM), Skudai, 81310, Johor Bahru, Johor, Malaysia
b
Department of Mechanical, Aerospace and Civil Engineering, Pariser Building, The University of Manchester, Manchester, United Kingdom
c
Department of Mechanical Engineering, The University of Melbourne, Parkville, VIC, 3010, Australia
d
Department of Engineering and Computing, National University, 11255 N Torrey Pines Rd, San Diego, CA 92037, United States
e
Buckinghamshire New University, High Wycombe Campus, Queen Alexandra Road, High Wycombe, Buckinghamshire, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: Fused Deposition Modelling (FDM), is an additive manufacturing technology where polymers are extruded using
Finite element analysis appropriate processing parameters to achieve suitable bonding while ensuring that overheating does not occur.
Artificial Neural Network Among processing parameters, polymer inlet temperature, nozzle size, extrusion speed, and air cooling speed are
Polymers
significantly effect on the extrusion process at the distance between the build plate and the nozzle tip (standoff
Additive manufacturing
3D printing
region). This study aims to evaluate the influences of the processing parameters on the thermal behavior and
phase change zone of Polyamide 12 (PA12) and Acrylonitrile Butadiene Styrene (ABS) polymers at standoff
region. A nonlinear three-dimensional (3D) finite element (FE) model was developed by implementing an
apparent heat capacity model using the Heat Transfer Module in COMSOL® Multiphysics software. FE results in
the standoff region were validated by experimental tests, concerning various nozzle sizes and extrusion speed.
The validated numerical results demonstrated that there is a complex correlation between processing parameters
and thermal behaviors such as phase change and temperature distribution in the standoff region. The FE results
were then employed in training an artificial neural network (ANN). A well-established compromise between the
trained ANN and the FE results demonstrates that the trained ANN can be employed in the prediction of further
thermal and glass transition behavior using subsequent processing parameters.

1. Introduction cost and is efficient in terms of material, waste, energy and production
time. Hence, FDM is one of the most popular additive manufacturing
Three-dimensional (3D) printing technologies such as Fused Depo­ techniques and is extensively used in automotive, aerospace, medical,
sition Modelling (FDM), selective laser sintering and, more generally, education and civil engineering industries [11–15]. In FDM, a feeder
additive manufacturing (AM) are of interest to researchers over the last introduces a polymer filament into a heating block, where it is melted to
few decades [1–3]. Unlike subtractive manufacturing technologies, ad­ a temperature higher than its glass transition temperature. Then, the
ditive manufacturing methods create 3D parts directly from 3D CAD molten polymer is extruded, cooled, and solidified onto the build plate
models by adding melted polymer material layer by layer. Subtractive layer by layer to form a 3D model. This method offers relatively poor
manufacturing methods typically use fabrication equipment such as accuracy with inferior surface quality when compared to models printed
cutting tools, fixtures and jigs [4]. In recent years, 3D printing tech­ using other manufacturing techniques in the AM category [16]. In
niques have become popular manufacturing processes and have been polymer 3D printing techniques such as FDM, wrapping, shrinkage,
recognized as contributing to the ‘third industrial revolution’ [5–10]. dimensional accuracy, and adhesion of printed material to the build
For small product quantities, FDM technology is significantly low- plate are primarily influenced by thermal factors such as extrusion

* Corresponding author.
E-mail addresses: [email protected] (N. Moslemi), [email protected] (Behzad. Abdi), [email protected] (S. Gohari), Izman@utm.
my (I. Sudin), [email protected] (E. Atashpaz-Gargari), [email protected] (N. Redzuan), [email protected] (A. Ayob), [email protected] (C. Burvill), meini.su@
manchester.ac.uk (M. Su), [email protected] (F. Arya).

https://doi.org/10.1016/j.applthermaleng.2022.118533
Received 23 November 2021; Received in revised form 8 April 2022; Accepted 13 April 2022
Available online 30 April 2022
1359-4311/© 2022 Elsevier Ltd. All rights reserved.
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 1. (a) The schematic of the extrusion system in FDM process, (b) heat transfer in an element of the molten polymer within the standoff region of length dx.

Fig. 2. A schematic of designed extrusion system and infrared thermal camera.

temperature, cooling rate, and melting temperature [17]. One solution quality of the deposited parts [18,22–25]. In other words, the quality of
to these temperature related issues is to determine best processing the 3D printed products in terms of the thermal and mechanical prop­
parameter values and then to calibrate the FDM to these processing erties are depended of the quality of the bonding between the lines and
parameters [18–21]. In FDM, the processing parameters include: layer layers which can be controlled by adopting optimized parameters.
thickness, build orientation, raster angle, air distance, extrusion tem­ Therefore, precise setting of processing parameters is critical to
perature, print speed, infill pattern, nozzle diameter, raster width, and achieving best FDM printing performance [7,26].
air-cooling rate influence on the mechanical properties and the final In the extrusion process (Fig. 1), the molten polymer at a certain

2
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

0.995

0.99

Non-Dimensional Temperature
0.985

0.98

0.975

0.97 Nozzle 0.4mm / Speed 4.4mm/s

0.965 Nozzle 0.4mm / Speed 11.3mm/s
Nozzle 0.4mm / Speed 20.2mm/s
0.96
Nozzle 0.6mm / Speed 4.9mm/s
0.955
Nozzle 0.6mm / Speed 8.2mm/s
0.95
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x (mm)

Fig. 3. The experimental results of temperature distribution of extruded ABS polymer at the distance between the nozzle tip and build plate.

Fig. 4. The temperature-dependent properties of PA12 polymer [70].

temperature is extruded from the nozzle tip and is then laid down over
Table 1 the build plate. Within the process, three heat transfer processes occur:
The thermal conductivity, density and specific heat capacity of PA12 and ABS in the nozzle, the distance between the nozzle tip and the build plate
polymer in both glassy and rubbery states [63]
(standoff region), and at the build plate. These heat transfer processes
Thermal properties Glassy state Rubbery state Unit affect the mechanical properties of the deposited part [27]. Despite
PA12 typically being a short distance, the temperature distribution of the
Thermal conductivity 0.078 0.32 W/(m.K) extruded polymer in the standoff region plays an important role in the
density 710 1060 kg/m3 process due to the associated unconstrained thermal gradients. Pro­
Heat capacity 2095 2215 J/(kg.K)
cessing parameters play a significant role in identifying the thermal
ABS
Thermal conductivity 0.3 0.2 W/(m.K) behavior of polymer in the standoff region.
density 1050 970 kg/m3 Several research studies have been conducted to evaluate the heat
Heat capacity 1200 1708 J/(kg.K) transfer process in the distance between the nozzle tip and build plate
using analytical, numerical, and experimental investigations
[17,28–30].
Heat transfer plays a significant role in the melting and solidification
of the polymer during the printing process. Controlling the heat transfer

3
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 5. FE simulation including: (a) boundary conditions for first step of meshing, (b-d) Adaptive mesh refinement in the extrusion process. (e-g) the results obtained
from b-d meshing models, respectively.

can significantly improve the quality of bonding of the layers which can experimentally using various techniques, for example, infrared ther­
be achieved by keeping the temperature of the printed lines greater than mography cameras [33–36]. Several experimental research studies have
its glass transition temperature [31,32]. The heat transfer process of the been conducted to define the heat transfer at the nozzle assembly
extruded polymer during the 3D printing process was determined [17,28,37]. During the 3D printing process of polymers, several heat

4
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

thermography imaging method. Their results show a 13% difference

between the numerical and experimental models. Zhou et al. [43]
concluded such discrepancy is due to inaccurate modelling of the build
plate heat transfer coefficient. Zhou et al. [43] also stated that the model
with the build plate temperature set at the highest level might require
further consideration. Improvement in the modified model can be used
for the prediction of thermal behavior. Using continuous media theory, a
thermal finite element simulation of the FDM 3D printing process of ABS
materials was developed by Zhou et al. [44] and the results showed by
changing the thermal properties of the material, the modelling method
would significantly influence the thermal evaluation.
Several research studies have reported on the characterization and
optimization of the 3D printing processing parameters [45–50]. For ABS
polymer, the effect of layer height, layer width and extrusion speed on
surface roughness of the FDM 3D printed parts was studied by Anitha
et al. [45]. The obtained results revealed that the effect of layer height is
most significant [45]. A derived relation was used as a cost function for
optimization [43]. To optimize the surface roughness of the FDM 3D
printed parts, Thrimurthulu et al. [46] developed an analytical model
Fig. 6. The architecture of a trained ANN in this study. using the genetic algorithm to predict the orientation of the 3D model
regardless of the complexity of the 3D model [51–53]. Four processing
parameters, including the printing speed at two different sections of the
transfer processes are present: radiation, convection, and phase change.
model, layer height and layer width were considered by Zhang et al.
All heat transfer processes significantly impact on the quality of the final
[47]. The relation between these parameters and the accuracy and
3D printed part. As such, the relative impact of these processing pa­
deformation of the FDM 3D printed parts were established using fuzzy
rameters needs to be determined. This requires high familiarity and
comprehensive evaluation and Taguchi method and the optimum pro­
accurate knowledge of the heat transfer process in the standoff region.
cessing parameters were determined. A statistical optimization method
Unlike metals, the low thermal conductivity of polymers limits
was used by Wang et al. [48] to find the optimum processing parameters
heating and cooling rates which results in shrinking and delamination
to optimize the surface roughness of the FDM 3D printed parts. A
during a cooling process [26]. Semi-crystalline polymers have a thermal
mathematical model was developed by Thrimurthulu et al. [46] to
conductivity higher than amorphous polymers. The structure of crystals
optimize the printing time. Thrimurthulu et al. [42] established the
in the polymer and consequently the physical properties of the semi-
build orientation as the effective processing parameter and the optimum
crystalline polymers are dependent on the rate of heat transfer during
build orientation was determined using the genetic algorithm.
the cooling process [38]. A high percentage of large crystals are formed
Artificial Neural Networks (ANNs) have been increasingly used as
in polymer during their slow cooling process which remarkably in­
non-parametric predictors for solving complex problems [54]. Some
creases on its hardness [27]. The 3D printing process is progressive,
researchers have used ANN to predict part quality in nonconventional
where a 3D model is built layer by layer. During this process, the
and conventional machining process such as milling and drilling
deposited and solidified layer or layers experience a heating effect, due
[55–57]. Their ANN results demonstrated that there is a close correla­
to subsequent layer(s), which influence the polymer’s crystallinity. This
tion between the processing parameters and the quality of the manu­
leads to uneven cooling and reheating processes which can introduce
factured part. The processing parameters of numerous conventional
residual stresses and as such, a reduction in the mechanical properties of
manufacturing technologies can be calibrated during 3D printing pro­
the 3D printed model. Prajapati et al. [34] observed that in FDM models
cess. While, adjusting some of the processing parameters such as print­
with thin walls, the extrusion temperature influences the extent of
ing pattern, printing orientation, layer thickness, nozzle diameter and
cracking and wrapping in the final model.
shell thickness during the operation is unmanageable during the FDM
FE simulation is extensively used as a prediction tool to evaluate the
3D printing process [58]. Therefore, to train an ANN, the off-line
cause and effect of the quality of deposited parts [22,39,40]. Thermo-
approach is chosen as a function of some processing parameters. Re­
mechanical analysis of FDM process was conducted by Courter et al.
searches have not thoroughly studied the combined effect of processing
[41] using FE simulation and the effect of tool path on the residual
parameters and their correlation with the final quality of the FDM 3D
stresses of the part of the model was investigated. Courter et al. [41]
printed parts [59–61]. The deficiency of a design method, especially in
conducted an FEA study in which the thermal properties of materials
FE, and the lack of a reliable prediction technique might be the reason
including thermal expansion coefficient and conductivity were consid­
for many associated problems. Systematically conducted robust FE,
ered constant with only the specific heat capacity of polymer being a
analytical methods as well as a trustworthy prediction methods are
function of temperature. Zhou et al. [41] observed that there is desirable
currently lacking.
interaction between the tool path and the thermos-mechanical proper­
The aim of this study is to evaluate the significance of the combined
ties of the 3D printed parts. Although good agreement between the
effect of processing parameters using developed FE simulation analysis
realistic and FEA results were observed, Zhou et al. [41] suggested that it
and an ANN Prediction Algorithm followed by experimental validation
is necessary to define all the thermal properties of polymers as a function
of the numerical results. The numerical and experimental results in this
of temperature and crystallinity. FE simulation and theoretical model­
research contribute to the investigation and prediction of the phase
ling of heat transfer and rheology of Poly lactic acid (PLA) polymer in
change zones and thermal behavior of ABS and PA12 polymers the
the nozzle were carried out by Go et al. [30] and Phan et al. [42]. Go
standoff region. To address the main objective of the study, a 3D tran­
et al. [30] and Phan et al. [42] defined the thermal expansion coefficient
sient finite element model was developed by integration of an apparent
as a function of temperature at two phases and reported the heat transfer
heat capacity method and adaptive mesh refinement method. Due to
between the PLA polymer and nozzle wall affect the overall printing
variation of thermal material properties of polymer by temperature, to
quality. Zhou et al. [43] developed a 3D finite element model for ther­
achieve accurate numerical results, the thermal material properties of
mal simulation of the FDM 3D printing process using birth-and-death
polymer were defined as a function of polymer temperature. The FE
feature. Zhou et al. [43] validated the model using the infrared
simulation was conducted using the Heat Transfer Module in COMSOL®

5
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

1 1

0.995
0.99
Non-Dimensional Temperature

Non-Dimensional Temperature
0.99

0.98
0.985

0.98
0.97
Experiment 0.975 Experiment

0.96 Analytical Analytical

0.97
FEM FEM
0.95 0.965
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x (mm) x (mm)

(a) (b)
1 1

0.998 0.998

0.996 0.996

Non-Dimensional Temperature
Non-Dimensional Temperature

0.994 0.994

0.992 0.992

0.99 0.99

0.988 0.988
Experiment Experiment
0.986 0.986 Analytical
Analytical
0.984 0.984 FEM
FEM
0.982 0.982
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x (mm) x (mm)

(c) (d)
1 1

0.995
0.995
Non-Dimensional Temperature
Non-Dimensional Temperature

0.99
0.99

0.985
0.985
0.98 Experiment
Experiment

Analytical 0.98 Analytical

0.975
FEM FEM

0.97 0.975
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x (mm) x (mm)

(e) (f)
Fig. 7. Comparison of FE simulation associated with the thermal behavior of polymer in the distance between the build plate and nozzle tip: (a) Nozzle size 0.4 mm
and speed 4.5 mm/s, (b) Nozzle size 0.4 mm and speed 7.5 mm/s, (c) Nozzle size 0.4 mm and speed 16.5 mm/s, (d) Nozzle size 0.4 mm and speed 20 mm/s, (e)
Nozzle size 0.6 mm and speed 6 mm/s, (f) Nozzle size 0.6 mm and speed 8 mm/s.

Multiphysics software package. The FE simulation results were then conductivity, heat capacity, and density are defined as a function of
validated by the experiment results. The influence of several processing temperature and the finite element modelling of the extrusion process in
parameters including: nozzle size, extrusion, and air cooling speed on the standoff region is discussed. Section 6 refers to the application of
the glass transition and thermal behavior of polymers was numerically ANN in the prediction of glass transition and thermal behavior of
studied. The FE simulation results were used to train an ANN for pre­ polymer in the distance between the nozzle tip and build plate. ANN
dicting and optimizing the processing parameters of the FDM 3D training is based on the FE simulation results. Finally, the FE simulation
printing. and ANN results are discussed in Section 7.
This paper is structured as follows: experimental procedure to
determine the temperature distribution of the extruded polymer at the 2. Experimental apparatus and test procedures:
distance between nozzle tip and build plate is presented in section 2 and
section 3. In Section 4, the mathematical formulation of heat transfer in To determine the temperature distribution of extruded polymer at
the distance between the nozzle tip and build plate is presented. In the standoff region, an infrared thermal camera was used. To provide a
Section 5, thermal material properties of polymers, including: thermal reliable optical access to the distance between the nozzle tip and build

6
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 8. The temperature distribution of ABS polymer in the distance between the nozzle tip and build plate: (a) Nozzle size 0.4 mm and speed 6 mm/s, (b) Nozzle size
0.6 mm and speed 8 mm/s.

Fig. 9. the variation of temperature in the transverse direction and within the cross-section of extruded polymer: (a) Nozzle size 0.4 mm and speed 6 mm/s, (b)
Nozzle size 0.6 mm and speed 8 mm/s.

plate, custom-made polymer extrusion system encompasses a set of E3D nozzle sizes 0.4 mm and 0.6 mm. Increasing the extrusion speed results
hotend (including: exchangeable nozzle, heater, Heater block, NTC100K in the extruded polymer losing less energy, i.e. increase the extrusion
thermistor, nozzle barrel and heat sink), filament feeder (including speed, decrease the temperature variation in the standoff region.
stepper motor, gear, and ball bearing) and an Arduino Mega 2560 and
ramps 1.6 based motherboard with stepper driver of DRV8850 was 4. Mathematical modelling
designed and used for accurate extrusion of polymer. Fig. 2 illustrates a
schematic of design extrusion system and infrared thermal camera. In the extrusion process, as illustrated in (Fig. 1a), the melted poly­
mer with a mass flow of ṁ leaves the nozzle tip (x = 0) and deposits on
3. Experimental results and discussion the build plate (x = L). As polymer is extruded from the nozzle tip, two
types of heat transfers including radiation and convection occurs at the
Upon accurate calibration of extrusion system, using ABS polymer distance between the nozzle tip and build plate. For simplification
and two different sizes of brass nozzle (0.4 mm and 0.6 mm), five ex­ purpose, it was assumed that the temperature distribution of polymer in
periments were conducted at different extrusion speeds and the tem­ this region is steady-state; and, the temperature of the build plate does
perature distribution along the standoff region (up to × = 1 mm from not have any effect on the thermal behavior of the extruded polymer.
the tip of the brass nozzle) were recorded. The obtained results are Using the mathematical relationship in (Fig. 1b) and considering the
presented in Fig. 3, which shows that in the standoff region, the highest thermal material properties of polymer and two types of heat transfers,
and the lowest temperature variations occur when the polymer is the total energy balance of an element of length dx can be derived as per
extruded using a nozzle size of 0.4 mm at an extrusion speed of 4.4 mm/s Eq.1 [62]:
and 20.2 mm/s, respectively. The extrusion speed has a significant in­
fluence on temperature variation within the standoff region for both

7
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 10. The overall energy flux of molten ABS polymer during the extrusion process.

( )
dT ( ) cooling, printing speed, and melt flow and glass transition regions
ṁCp T = ṁCp T + dx + hc (T − Tam )Pdx + ∈ σ T 4 − Tr4 Pdx (1)
dx should be considered when developing an accurate model [63].
Heat transfer is then derived based on the assumption in [60] as Eq.6:
where, P, ṁ and Cp stand for the diameter of extruded polymer, mass
ρCp u.∇T + ∇.( − k∇T) = Q (6)
flow rate, and the heat capacity of polymer, respectively. Tam , ∈, σ and hc
are the ambient temperature, emissivity of polymer, Stefan-Boltzmann
where,Cp , Q, k, and u are the specific heat capacity of polymer, heat
constant, and convective heat transfer coefficient, respectively.
power per volume unit, thermal conductivity, and extrusion speed in
The total energy balance equation (Eq.1) can be defined in a brief
two states, respectively. After polymer is extruded from the nozzle, it
format as per Eq.2 [62]:
cools to its glass transition state and releases a large amount of energy.
dT hc P ∈ σP ( 4 ) Considering the changes in heat capacity and enthalpy, the amount of
= − (T − Tam ) − T − Tr4 (2)
dx ṁCp ṁCp energy per unit mass can be calculated using Eq.7 which is based on the
variation of the specific heat capacity [63]:
Due to the non-linearity of Eq.2, defining an analytical solution is
very challenging. The numerical simulation of Eq.2 can be derived by ΔCp =
ΔH
(7)
linearizing the radiative term through writing an estimated radiative T
heat transfer coefficient, hr (Eq.3) [62]: To relate the heat to glass transition of polymer, an apparent heat
( )( ) capacity method is used for the heat transfer with a phase change
2
hr =∈ σ Teff + Tr2 Teff + Tr (3)
domain [64,65]. The aim of this analysis is to define ΔT and by
considering that the flow is laminar, the fluid velocity (u) and pressure
The effective temperature of polymer (Teff ) can be estimated by
(p) are stated in Eqs.8,9 [63]:
average temperature of extruded polymer which can be calculated using
[ ( ) ]
Eq.4: ∂u 2μ
ρ + ρu.∇u = ∇. − pI + μ(∇u) + (∇u)T − − κ (∇.u)I + F (8)
dT heff P ∂t 3
= − (T − T∞ )heff = hc + h∞ (4)
dx ṁCp
∂ρ
+ ∇.(ρu) = 0 (9)
Eq.5 is used to calculate the temperature distribution in polymer: ∂t
T − T∞
(
x
)
ṁCp where, κ and μ represent the dilatational viscosity and viscosity of
θ= = exp − x0 = (5) polymer, respectively. The term F is the dampen of the velocity of
T0 − T∞ x0 heff P
polymer at the modulus-change interface, which can be calculated using
Based on Eq.5, the temperature of polymer reduces nonlinearly along Eq.10 [63]:
the extrusion direction [16].
Using fluid dynamic and thermal analysis, the conduction behavior (1 − α)2
F= A (u − uex ) (10)
of polymer were defined by Rahman et al. [63]. Other processing pa­ α3 + ε m
rameters in the polymer extrusion process including: the nozzle size, air

8
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 11. Thermal behavior of ABS and PA12 Polymers at the distance between the nozzle tip and build plate: (a) ABS polymer (speed 4 mm/s), (b) ABS polymer
(speed 10 mm/s), (c) PA12 polymer (speed 4 mm/s), (d) PA12 polymer (speed 10 mm/s).

Fig. 12. The glass transition behavior of polymers using a nozzle with diameter 0.4 mm: (a) ABS polymer at speed 0.15 mm/s, (b) PA12 polymer at speed 0.26 mm/s.

Where, uex , α, ε and Am are the velocity, volume fraction of the liquid such as temperature, air cooling and extrusion speed. Therefore, a
phase and arbitrary constants, respectively [28]. converged mesh in the phase change zones should be employed to assess
The glass transition behavior of polymers in the FDM 3D printing accurate results.
process is nonlinear and it is a function of various processing parameters

9
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Table 2
The inlet speed of polymer in extrusion system at a constant outlet speed of 4 mm/s.
Nozzle size (mm) 0.1 0.2 0.3 0.35 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Inlet velocity (mm/s) 0.0131 0.052 0.117 0.16 0.209 0.326 0.47 0.64 0836 1.06 1.31

1 1

0.9 Nozzle 0.9 Nozzle

0.1mm 0.1mm

Non-Dimensional Temperature
Non-Dimensional Temperature

0.8 0.8
0.2mm 0.2mm
0.7 0.7
0.3mm 0.3mm
0.6 0.6
0.4mm 0.4mm
0.5 0.5mm 0.5 0.5mm
0.4 0.6mm 0.4 0.6mm
0.3 0.7mm 0.3 0.7mm

0.2 0.8mm 0.2 0.8mm

0.9mm 0.9mm
0.1 0.1
1.0mm 1.0mm
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x.mm x.mm

(a) (b)
Fig. 13. The effect of nozzle size on non-dimensional temperature of the polymer at the distance between the nozzle tip and build plate: (a) ABS polymer, (b)
PA12 polymer.

Fig. 14. The glass transition behavior of ABS and PA12 polymers using a nozzle with diameter of 0.3 mm at extrusion speed of 0.117 mm/s and air-cooling speed of
15 m/s: (a) ABS polymer, (b) PA12 polymer.

5. Finite element simulation: semi-crystalline polymer [69] and Acrylonitrile Butadiene Styrene
(ABS) as an amorphous polymer [3] selected and their thermal material
5.1. Material modeling properties including the temperature-dependent heat capacity, heat
conductivity and density of polymers were obtained from the literature
The thermal properties of the polymer including the glass transition [63,70,71] which is illustrated in Fig. 4. The melting and solidification
temperature, the melting temperature, the heat capacity and the thermal temperature of PA12 polymer is at 169.7 ◦ C and 154.2 ℃, respectively
conductivity have considerable effect on the overall quality of 3D [70]. The thermal conductivity, density and specific heat capacity of
printed [66]. The melting temperature of semi-crystalline polymers ABS and PA12 polymers in both glassy and rubbery states at constant
occurs at a sharp point, while amorphous polymers do not have a spe­ pressure are listed in Table.1 [63].
cific melting temperature. In other words, they are high viscos in the
melt state which results in no flowability since the viscosity is reduced 5.2. Finite element modeling
by increasing the temperature [67]. Therefore, the peak of the differ­
ential scanning calorimetry (DSC) curve is assumed to be the melting To simulate the polymer phase change, the Heat Transfer Module in
temperature of amorphous polymer[68]. COMSOL® Multiphysics software package (based on an apparent heat
In this study, the Polyamide 12, PA12 (PA2200, EOS GmbH) as a capacity method) was used. This Module is suitable for various phase

10
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 15. Thermal behavior of ABS polymer at the distance between the build plate and nozzle tip at different setup: (a) Nozzle size 0.4 mm and speed 3 mm/s, (b)
Nozzle size 0.4 mm and speed 5 mm/s, (c) Nozzle size 0.4 mm and speed 7 mm/s, (d) Nozzle size 0.8 mm and speed 3 mm/s, (e) Nozzle size 0.8 mm and speed 5 mm/
s, (f) Nozzle size 0.8 mm and speed 7 mm/s.

transitions [72,73]. To use this COMSOL® Module for accurate model­ effect of temperature-dependent material properties. A continuous
ling of phase change between the solid and liquid states, the thermal adaptive mesh refinement algorithm in COMSOL® was used to find the
material properties of polymer should be defined separately for both solidification region. Based on this algorithm, the simulation is initiated
states as a function of temperature and phase transition function α(T). by a gradual transition between the glassy and rubbery states using
Properties include the heat capacity, thermal conductivity, and density coarser mesh in first stage. The first simulation was then used as a
of polymer. The phase transition function is equal to zero and one for a starting point for the next step of simulation with a finer mesh, and this
pure solid and a pure liquid, respectively [60]. In the FE simulation, 2D process was repeated until the best convergence was achieved. In the
axisymmetric models, due to the extrusion system’s axial symmetry, present study, the linear trial mesh was used for meshing the model. The
were used. Parameters that significantly affect the location of transition boundary condition for the first step of meshing is illustrated in (Fig. 5a
between the glassy and rubbery states include: extrusion speed, cooling and Fig. 5b-g), and demonstrates the adaptive mesh refinement method.
speed, nozzle size and extrusion temperature (Section 1). Therefore, it is The boundary conditions in (Fig. 5a) are similarly adopted for the next
necessary to use fine mesh across the transition state to consider the steps of meshing model.

11
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 16. Thermal behavior of PA12 polymer at the distance between the build plate and nozzle tip at different setup: (a) Nozzle size 0.4 mm and speed 3 mm/s, (b)
Nozzle size 0.4 mm and speed 5 mm/s, (c) Nozzle size 0.4 mm and speed 7 mm/s, (d) Nozzle size 0.8 mm and speed 3 mm/s, (e) Nozzle size 0.8 mm and speed 5 mm/
s, (f) Nozzle size 0.8 mm and speed 7 mm/s.

6. The artificial neural network approach: other input data that has not been used during the training process. This
capability is called generalization [77,78].
Artificial neural networks (ANNs) are non-parametric predictors that ANNs consist of four steps including the definition of input and
offer useful and viable means for solving complex problems and for outputs, setting database, designing, and training the network and
solution predictions [54]. ANNs can predict hidden functional re­ evaluation. The aim of using ANNs in this study is to determine the
lationships between the effective processing parameters and polymer temperature and phase change rate in the extrusion process. A set of
phase change or thermal behavior of the extrusion system as inputs and effective processing parameters as discussed before including the inlet
output of networks, respectively [74]. ANNs were developed to mimic temperature and speed, nozzle diameter, cooling rate and type of ma­
the biological nervous system [75,76]. The nodes or neurons, weights terial are ANN input parameters. The temperature and phase change rate
and connections (dendrites and synapses) are the main components of at different points along the distance between the nozzle tip and build
ANNs. A set of input and out data should be used to train an ANN. Once plate are as output ANN parameters. A set of the normalized database
the training of an ANN is done, the ANN can then be used to estimate consists of 98 finite element models were analyzed and used in

12
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 17. The phase change behavior of ABS polymer at two air-cooling speeds including (a) 1.5 m/s, (b) 7 m/s.

Fig. 18. The two-phase transition between the liquid and solid phases of PA12 polymer extrusion at two different air-cooling speeds: (a) 1.5 m/s and (b) 7 m/s.

training different architecture of ANNs using the Levenberg-Marquardt offers the highest performance for this study (Fig. 6).
[58,79–81] algorithm in MATLAB Software. From that normalized The inlet temperature and speed, air cooling rate, types of materials
database, 80% of data was used for training the networks and 15% and and nozzle diameter were considered as input parameters for the ANN
5% of data was used for validation and performance testing, respec­ and the temperature and phase change rate at different points alongside
tively. After compiling the database, the integrity of data was checked the distance between the nozzle tip and build plate were assumed as
using Microsoft Excel Spreadsheet to find the most effective parameters output parameters (Fig. 6). The training process ended after 245 epochs
and eliminate the data which can make noise during the training with the mean error and validation set of 2.235e-6 and 4.857e-6,
process. respectively.
The performance of ANNs significantly depends on the architecture
of the network. As such, there is no guideline for the choice of appro­ 7. Results and discussions
priate architecture. Trial and error experiments are normally used to
assess the most-optimum ANN architecture. In this study, multilayer 7.1. Validation of the FE simulation results
perceptron architecture using the Levenberg-Marquardt algorithm and
trial and error method was used to test and find the optimum number of The FE simulation was carried out to determine the influence of four
hidden layers for the ANN with highest performance. It has been different inlet extrusion speeds 4.5 mm/s, 7.5 mm/s, 16.5 mm/s and 20
observed that an ANN consists of one input layer with five neurons, one mm/s for the nozzle size 0.4 mm and two different inlet extrusion speeds
output layer with five neurons, and one hidden layer with seven neurons 6 mm/s and 8 mm/s for nozzle size 0.6 mm. several points along the

13
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

Fig. 19. Network output vs real output for (a) train data b) test data.

surrounding environment and the nozzle surface.

Table 3 Fig. 8 shows the temperature distribution of ABS polymer in the
Comparison between the FE simulation results and the predicted ANN results. distance between the nozzle tip and build plate for nozzle sizes of 0.4
Model Model Model Model mm and 0.6 mm with extrusion speeds 6 mm/s and 8 mm/s, respec­
1 2 3 4 tively. Fig. 9 shows the thermal behavior of ABS along the cross- section
Extrusion speed (mm/s) 4 5 6 10 of extruded polymer for two different nozzle sizes of 0.4 mm and 0.6 mm
Nozzle diameter (mm) 0.4 0.9 0.45 0.65 with extrusion speeds 6 mm/s and 8 mm/s, respectively. It can be
Air cooling speed (m/s) 10 0 5 0 observed from Fig. 8 and Fig. 9 that the variation of temperature in the
Type of material ABS ABS PA12 PA12 transverse direction and within the cross-section is minimal, about 2%,
Inlet temperature (Celsius) 195 245 228 210
Non-Dimensional ANN: 0.983 0.998 0.993 0.99
and these results are almost the same as the recorded temperature
Temperature at x = FEA: 0.982 0.996 0.994 0.993 through the experiment. Comparing the present FEM results with the
0.3 mm Square 0.0001 0.0004 0.0001 0.0009 experimental and analytical results, demonstrates a reliable agreement
Error (%): in the temperature results in x and transverse directions. Therefore, the
Non-Dimensional ANN: 0.953 0.99 0.985 0.99
developed finite element model can be employed to study the influence
Temperature at x = FEA: 0.953 0.991 0.985 0.989
0.6 mm Square 0 0.0001 0 0.0001 of the processing parameters on the temperature distribution of the
Error (%): polymer in the extrusion system.
Non-Dimensional ANN: 0.922 0.984 0.974 0.984 The glass transition behavior of ABS polymer is evaluated at an
Temperature at x = FEA: 0.924 0.986 0.973 0.983 extrusion temperature of 105 ◦ C. Using a nozzle with a diameter of 0.4
0.9 mm Square 0.0004 0.0004 0.0001 0.0001
Error (%):
mm, the molten ABS polymer is extruded at a speed of 1 mm/s. In this
Phase change rate at x ANN: 0.998 1 0.999 0.998 study, the heat transfer (convection) between the surface of the nozzle
= 0.4 mm FEA: 1 1 1 1 and the surrounding environment (25 ◦ C) is considered. Fig. 10 shows
Square 0.0004 0 0.0001 0.0004 the overall energy flux in the process, where the lowest and highest
Error (%):
energy flux are released outside the nozzle and in the nozzle assembly,
Phase change rate at x ANN: 0.983 0.993 0.991 0.995
= 0.8 mm FEA: 0.985 0.994 0.993 0.994 respectively. The energy flux is reduced with distance from the nozzle
Square 0.0004 0.0001 0.0004 0.0001 center. A small amount of energy flux is released at the nozzle tip due the
Error (%): released heat in the secondary glass transition once the molten polymer
is extruded through the nozzle.
distance between the nozzle tip and build plate were identified and the
7.2. Nozzle size:
temperature distribution at these points was compared with experiment
results, and shown in Fig. 7. From the analytical and experiment results,
Using the FE simulation, the thermal behavior of two types of
it is observed that there is a nonlinear relationship between the tem­
polymers including ABS and PA12 polymers at the extrusion tempera­
perature of extruded polymer and the extrusion speed. Temperature is
ture of 215 ◦ C were studied. The effect of nozzle size on temperature
dropped when the distance from nozzle tip and extrusion speed is
distribution at two constant speeds (4 mm/s and 10 mm/s) and the
increased.
ambient temperature of 25 ◦ C, were analysed and the results are illus­
Comparing the FE simulation, experimental, and analytical results
trated in Fig. 11.
(Fig. 7), it can be seen that FEM results of the temperature of the
In this section, the effect of inlet speed was studied without consid­
extruded polymer estimate slightly lower and higher, respectively, in the
ering the impact of nozzle diameter on outlet speed. It can be seen from
first and second half of the distance between the build plate and nozzle
Fig. 11 that at a constant inlet extrusion speed, the nozzle size signifi­
tip. Low-temperature prediction in the first half can be attributed to two
cantly affects temperature distribution in the polymer. The temperature
assumptions: First, the nozzle surface is isolated and there is no heat
is reduced by increasing the nozzle size. Next, the increasing the size of
transfer between the environment and the nozzle surface; second, the
the nozzle from 0.2 mm to 0.6 mm at a constant speed of 4 mm/s will
extruded polymer is fully circular, and its diameter is constant. In reality
increase the temperature of ABS polymer and PA12 polymer at x =
and according to die-swell theory [42], the cross-section of polymer in
0.5mm by 3.2% and 1.84%, respectively. In other words, the effect of
the distance between the nozzle tip and build plate changes slightly and
nozzle size in the amorphous polymer (ABS) is higher than the semi-
there is some radiation and convection heat transfer between the
crystalline polymer (PA12). By increasing the size of the nozzle, the

14
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

heat transfer (convention) from the outer surface of the extruded poly­ temperature (25 ◦ C) were kept constant in all simulations. The obtained
mer is increased. In contrast, the temperature in the polymer’s center- results at different air-cooling speeds for ABS and PA12 are illustrated in
line is changed slightly because the cooling rate is higher on the outer Figs. 15 and 16, respectively.
surface. According to physics laws (the equation of continuity), Fig. 15 and Fig. 16 show that the effect of air-cooling on the tem­
increasing the diameter of the nozzle at a constant inlet speed reduces perature of the extruded for ABS and PA12 polymers is lower than other
the outlet speed which consequently influences the temperature distri­ processing parameters due to the short distance between the build plate
bution of the extruded polymer. Comparing the obtained results for the and nozzle tip. For instance, by adding an air-cooling process at the
temperature distribution of ABS polymer at two different extrusion speed of 7 m/s to an extrusion process using two sets of the nozzle
speeds (4 mm/s and 6 mm/s), indicated that the extrusion speed slightly including 0.4 mm and 0.8 mm, the temperature of the extruded polymer
affects the center-line temperature. For an example, in extrusion process at the middle of the distance between the nozzle tip and build plate is
of ABS polymer through two nozzle sizes of 0.5 mm and 0.8 mm, reduced by 0.5% and 0.05%, respectively. As mentioned in Subsection
increasing the speed from 4 mm/s to 10 mm/s increases the temperature 6.1, cooling processes are mainly occurred on the outer surface of the
at × = 0.5 mm by 0.8% and 0.15%, respectively. extruded polymer. The temperature variation between the outer surface
To study the glass transition behavior of polymer during the extru­ and the middle are 1.2% and 1%, respectively. Comparing graphs (a)
sion process, the extrusion process of ABS and PA12 polymers were and (c) in Fig. 15 shows that the effect of air-cooling speed on the
simulated at the same extrusion temperature (215 ◦ C) with various thermal behavior of the extruded ABS polymer at higher speed is lower
extrusion speeds as well as the effect of nozzle size on glass transition because, increasing the extrusion speed reduces the heat transfer’s
behavior. It was observed that no glass transition occurs in the distance duration in the mentioned region. For instance, adding an air-cooling
between the nozzle tip and build plate for both polymers. Under system at speed of 15 m/s to an extrusion process of PA12 polymer
respective extrusion conditions, the polymer remains liquid at the dis­ using two sets of nozzles including 0.8 mm and 0.4 mm at the extrusion
tance between the build plate and nozzle tip. It was noticed that the speed of 3 mm/s, reduces the temperature of the extruded polymer at
extrusion speed and temperature will affect the location of the glass the middle of the distance by 0.15% and 0.85%, respectively. Comparing
transition. For instance, the glass transition takes place in the distance the obtained results for both types of the polymers at different air-
between the build plate and nozzle if the extrusion speed of ABS polymer cooling rates showed that the effect of the air-cooling process on the
is set to lower than 0.65 mm/s and 0.3 mm/s for nozzle size of 0.2 mm temperature of the semi-crystalline polymer (PA12) is higher than the
and 0.4 mm, respectively. At the same conditions, the extrusion speed amorphous polymer (ABS) which should be assumed to set the pro­
for PA12 polymer should be lower than 1.1 mm/s and 0.56 mm/s, cessing parameters.
respectively. As illustrated in Fig. 12, the glass transition of ABS and The phase change between the solid and liquid states in the extrusion
PA12 polymers using a nozzle with a diameter of 0.4 mm occurs at the process of ABS and PA12 polymers at two different air-cooling speeds
distance between the nozzle tip and build plate if the extrusion speed is including 1.5 m/s and 7 m/s using a nozzle size of 0.1 mm and extrusion
set to 0.15 mm/s and 0.26 mm/s, respectively. speed of 1 mm/s are illustrated in Fig. 17 and Fig. 18, respectively. In
In practice, during the 3D printing process, the extrusion speed de­ this study, the extrusion processing parameters were chosen as the phase
pends on the printing speed. For optimization and better control of the change appears in the distance between the nozzle tip and build plate.
overall quality of the 3D printed model, it is crucial to have a thorough Comparing the obtained results, showed that the air-cooling process
understanding of the nozzle effects on temperature distribution and affects slightly on the location of phase change in the distance between
glass transition behavior of polymer at a constant speed and flow rate. the nozzle tip and build plate. The polymer phase is changed at air-
The outlet extrusion speed was kept constant at 4 mm/s, and the inlet cooling speed of 1.5 m/s is appeared at a distance of 0.35 mm from
extrusion speed for different size of nozzle was calculated using a the nozzle tip. By increasing the air-cooling speed to 7 m/s, the location
physics law which states that the volume flow rate at a nozzle is con­ of phase change is appeared at the distance of 0.23 mm from the nozzle
stant, and it is a function of cross-section and the speed of flow tip, which is closer to nozzle tip.
(Q = Ai vi = Ao vo ). Using volume flow rate law at a constant outlet speed
or polymer, the inlet speed of polymer is calculated and listed in Table.2. 7.4. Prediction of temperature distribution and phase change rate:
Next, the influence of nozzle diameter on glass transition and ther­
mal behavior of two types of polymers including PA12 and ABS poly­ As mentioned in Section 5, Five processing parameters, including the
mers were studied numerically at a constant extrusion temperature of nozzle diameter (0.1 mm to 1 mm), air cooling rate (0 to 15 m/s), type of
215 ◦ C, ambient temperature of 25 ◦ C and various inlet speed as listed in material (ABS, PA12), inlet temperature (180 ◦ C to 250 ◦ C) and extru­
Table 2. The obtained results are illustrated in Fig. 13. sion speed (3 mm/s to 10 mm/s) were set as inputs (Section 5). The glass
For both types of polymers, by keeping the outlet extrusion speed as a transition rate from rubbery state to solid-state and temperature at
constant value, it can be seen from Fig. 13 that the diameter of the nozzle different points along the distance between the nozzle tip and build plate
significantly affects the temperature distribution in the distance be­ were set as outputs. The inputs and outputs were then used to train the
tween the nozzle tip and build plate. By increasing the diameter of the network. (Fig. 19a and 19b) demonstrate the plot of network output vs
nozzle from 0.2 mm to 0.6 mm, the temperature of the polymer at × = real data for train data and test data of temperature and phase change
0.5 mm is increased by 88.2% and 73.5%, respectively. A large release of rate at different places along the distance between the nozzle tip and
energy in the small nozzles causes a complete phase change in build plate, respectively. As illustrated in the figures, the errors are low,
mentioned distance. A typical example is, for nozzle size of 0.3 mm, the and the output data is satisfactorily dispersed. This indicates that the
glass transition of ABS is occurred near the nozzle tip, and for PA12, it is ANN is readily able to predict temperature and phase change rates. It can
taken place far from the tip of the nozzle, as seen in Fig. 14. be also observed that the randomness of data was assured.
The accuracy of the trained ANN model and comparison its result
7.3. Air- cooling speed: with FE simulation results were also investigated using the four new
models which have not been used in the earlier network training pro­
To study the effect of air-cooling speed on the glass transition cesses. Each model contains different processing parameters, as shown
behavior and temperature distribution of extruded ABS and PA12 in Table 3. It can be seen in Table 3 that there is a close agreement
Polymers, two sets of nozzles including 0.4 mm and 0.8 mm were between the predicted results using trained ANN and FE results. For
selected and the extrusion process of these polymers at different extru­ instance, for ABS material at an extrusion speed of 5 mm/s and air
sion speeds 3 mm/s, 5 mm/s and 7 mm/s were simulated using devel­ temperature of 245 ◦ C, nozzle diameter of 0.9 mm without any air-
oped FE. The extrusion temperature (215 ◦ C) and the ambient cooling process, the square error between the ANN and FE results at

15
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

× = 0.6 mm is 0.1%. The maximum difference between the ANN and FE References:
results is 0.3% for temperature and 0.2% for phase change rate. Due to
the number of various processing parameters and interaction among [1] Apparatus for production of three-dimensional objects by stereolithography,
(August 1984).
them, the setting suitable processing parameters to successfully achieve [2] J.S.H. Emanuel M. Sachs, Michael J. Cima, Paul A. Williams, Three-dimensional
3D printing process is required. printing techniques, US Patent, 5,204,055 (1993).
The processing parameters are manually adjusted by user in most of [3] P. Wang, B. Zou, S. Ding, Modeling of surface roughness based on heat transfer
considering diffusion among deposition filaments for FDM 3D printing heat-
the slicing software in the FDM 3D printing technology. For those users resistant resin, Appl. Therm. Eng. 161 (2019), 114064.
who cannot access to FE for analyzing the effect of adjusted parameters [4] I. Gibson, D. Rosen, B. Stucker (Eds.), Additive Manufacturing Technologies,
on the quality of 3D printed models, the trained ANN can be integrated Springer New York, New York, NY, 2015.
[5] R.S.G.D.T. Pham, A comparison of rapid prototyping technologies, Int. J. Mach.
with a software slicing engine. In other words, implemented slicing Tool Manufact 38 (1998) 1257–1287.
software can predict the thermal behavior of selected polymer based on [6] W.E. Frazier, Metal additive manufacturing: a review, J. Mater. Eng. Perform 23
the user defined parameters to suitably define processing parameters. (6) (2014) 1917–1928.
[7] M.C.L.J.P. Kruth, T. Nakagawa, Progress in additive manufacturing and rapid
Such integration can guide FDM 3D printer users in smart processing
prototyping, CIRP Ann. 47 (1998) 525–540.
parameters selection and regulation. This procedure can be then alter­ [8] V.H.G.J.P. Vojislav, J.F. Olga, D.G. Javier, R.B.P. Jose, P.G. Luis, Additive layered
natively used in order to predict the temperature distribution and phase manufacturing: sectors of industrial application shown through case studies, Int. J.
change rate of polymer extrusion process in the 3D printing process. Prod. Res 49 (2010) 1061–1079.
[9] N. Li, S. Huang, G. Zhang, R. Qin, W. Liu, H. Xiong, G. Shi, J. Blackburn, Progress in
additive manufacturing on new materials: a review, J. Mater. Sci. Technol. 35 (2)
8. Conclusions (2019) 242–269.
[10] L.E. Murr, Frontiers of 3D Printing/Additive Manufacturing: from Human Organs
to Aircraft Fabrication, J. Mater. Sci. Technol. 32 (2016) 987–995.
The present study investigated the influence of the FDM 3D printing [11] A. Haleem, M. Javaid, 3D printed medical parts with different materials using
processing parameters, the inlet and outlet extrusion speeds, nozzle size, additive manufacturing, Clinical Epidemiology and Global, Health 8 (1) (2020)
and cooling rate, on the thermal and phase change behavior of ABS and 215–223.
[12] A. Bhardwaj, S.Z. Jones, N. Kalantar, Z. Pei, J. Vickers, T. Wangler, P. Zavattieri,
PA12 polymers. The analysis focused on the distance between the nozzle N. Zou, Additive Manufacturing Processes for Infrastructure Construction: a
tip and build plate in the FDM extrusion process using the steady state FE Review. Proceedings of the ASME 2019 14th International Manufacturing Science
simulation based on an apparent heat capacity method in COMSOL® and Engineering Conference. Volume 1: Additive Manufacturing; Manufacturing
Equipment and Systems; Bio and Sustainable Manufacturing, 2019.
Multiphysics software. The FE simulation results were compared with, [13] A.B. Davin Jankovics, D. Jankovics, A. Barari, Customization of Automotive
and validated by, a series of the experiments. The FE simulation results Structural Components using Additive Manufacturing and Topology Optimization,
were also used to train an ANN for further prediction. The FE simulation IFAC-PapersOnLine 52 (2019) 212–217.
[14] F.F. Ramdhani, B. Mulyanti, Additive manufacturing in education, IOP Conf.
results indicated that the thermal behavior of polymer and the
Series, Materials Science and Engineering 830 (4) (2020) 042093.
phase change rate of polymer in the distance between the nozzle tip [15] G.R. Manish Kamal, Design for metal additive manufacturing for aerospace
and build plate (standoff region) are influenced by extrusion processing applications, (2019) 67-86.
parameters. It was found that the nozzle size is one of the most effective [16] H. Bikas, P. Stavropoulos, G. Chryssolouris, Additive manufacturing methods and
modelling approaches: a critical review, Int J Adv Manuf Technol 83 (1-4) (2016)
processing parameters in the FDM 3D printing process. The temperature 389–405.
of extruded polymer in the distance between the nozzle tip and build [17] M.E. Mackay, Z.R. Swain, C.R. Banbury, D.D. Phan, D.A. Edwards, The
plate was reduced by increasing the nozzle size, which effects the performance of the hot end in a plasticating 3D printer, J. Rheol 61 (2) (2017)
229–236.
amorphous polymer (ABS) far more than the semi-crystalline polymer [18] A. Jaisingh Sheoran, H. Kumar, Fused Deposition modeling process parameters
(PA12). Furthermore, it was observed that the extrusion speed effects optimization and effect on mechanical properties and part quality: review and
the glass transition zone along the distance between the nozzle tip and reflection on present research, Mater. Today:. Proc. 21 (2020) 1659–1672.
[19] C. Abeykoon, P. Sri-Amphorn, A. Fernando, Anura Fernando, Optimization of fused
build plate and the center-line temperature. Due to the short distance deposition modeling parameters for improved PLA and ABS 3D printed structures,
between the build plate and nozzle tip, air cooling rate on the temper­ International Journal of Lightweight Materials and Manufacture 3 (3) (2020)
ature of the extruded polymers is lower than that of other processing 284–297.
[20] S. Wang, Y. Ma, Z. Deng, S. Zhang, J. Cai, Effects of fused deposition modeling
parameters and its effect on the temperature of the semi-crystalline process parameters on tensile, dynamic mechanical properties of 3D printed
polymer (PA12) is higher than that of the amorphous polymer (ABS). polylactic acid materials, Polym. Test. 86 (2020).
The comparison of the numerical and experimental results showed [21] T.N.H.V.H. Nguyen, T.P. Nguyen, T.T. Tran, Single and Multi-objective
Optimization of Processing Parameters for Fused Deposition Modeling in 3D
good agreement which demonstrates that FE simulation can be
Printing Technology, international journal of automotive and mechancial
employed to train an ANN. In addition, the trained ANN can be adopted engineering 17 (2020) 7542–7551.
to further predict the phase change rate and thermal distribution of [22] S.K.P. Ashutosh Pandey, Investigations into Complete Liquefier Dynamics and
polymers at the distance between the nozzle tip and build plate. Both FE Optimization of Process Parameters for Fused Deposition Modeling, Mater. Today:.
Proc. 5 (2018) 301–307.
simulation and ANN methods are reliable and robust methods and can [23] S.H.M.O.A. Mohamed, J.L. Bhowmik, Optimization of fused deposition modeling
be used to optimize the processing parameters in the FDM 3D printing. process parameters: a review of current research and future prospects, J. Adv.
This study also demonstrated that a trained ANN can be used as an Manuf. 3 (2015) 42–53.
[24] N.Y.A. Dey, A systematic survey of FDM process parameter optimization and their
alternative to FE. influence on part characteristics, J. Manuf. Mater. Processing 3 (2019) 64.
[25] S.M.M. Moradi, A. Kaplan, 3D printed parts with honeycomb internal pattern by
Declaration of Competing Interest fused deposition modelling; experimental characterization and production
optimization, Met. Mater. Int. 25 (2019) 312–1325.
[26] C. Rauwendaal, Polymer extrusion, Hanser Publishers, München, 2001.
The authors declare that they have no known competing financial [27] G.R. Stroble, The Physics of Polymers, Springer, Berlin, 1997.
interests or personal relationships that could have appeared to influence [28] B.D.V.F. Peng, M. Cakmak, Complex flow and temperature history during melt
extrusion in material extrusion additive manufacturing, Addit. Manuf 22 (2018)
the work reported in this paper. 197–206.
[29] S.G.A. Bellini, M. Bertoldi, Liquefier dynamics in fused deposition, J. Manuf. Sci.
Acknowledgements Eng 126 (2007) 237–246.
[30] A.J.H.J. Go, Fast desktop-scale extrusion additive manufacturing, Addit. Manuf 18
(2017) 276–284.
N.M would like to acknowledge the funding received from UTM [31] B.G. Compton, B.K. Post, C.E. Duty, L. Love, V. Kunc, Thermal analysis of additive
under the Post-Doctoral Fellowship Scheme (Grant. No. Q. manufacturing of large-scale thermoplastic polymer composites, Addit. Manuf. 17
(2017) 77–86.
J130000.21A2.05E30) for the Project: “Uniaxial and Biaxial Ratcheting
[32] F.M.D.S.F. Costa, J.A. Covas, Estimation of filament temperature and adhesion
of a Girth-Welded Super Duplex Stainless Steel (UNS S32750) Pressur­ development in fused deposition techniques, J. Mater. Process. Technol 245 (2017)
ized Pipe”. 167–179.

16
N. Moslemi et al. Applied Thermal Engineering 212 (2022) 118533

[33] D.W.D. Anthony, D. Wetz, A. Jain, Non-invasive measurement of internal [56] J. Balic, Neural-network based numerical control milling machine, J Intell Robot
temperature of a cylindrical Li-ion cell during high-rate discharge, Int. J. Heat Mass Syst 40 (2004) 343–358.
Tran 111 (2017) 223–231. [57] P. Muñoz-Escalona, P.G. Maropoulos, Artificial neural networks for surface
[34] Y.A.L. Dupont, P. Jeannin, Comparison of junction temperature evaluations in a roughness prediction when face milling Al 7075–T7351, J. Mater. Eng. Perform. 19
power IGBT module using an IR camera and three thermosensitive electrical (2) (2010) 185–193.
parameters, Int. J. Heat Mass Tran 49 (2013) 1599–1608. [58] V.G.A. Boschetto, F. Veniali, Surface roughness prediction in fused deposition
[35] H.F. Hamann, A. Weger, J.A. Lacey, Z. Hu, P. Bose, E. Cohen, J. Wakil, Hot spot modelling by neural networks, Int J Adv Manuf Technol 67 (2013) 2727–2742.
limited microprocessors: direct temperature and power distribution measurements, [59] R.V. Pazhamannil, P. Govindan, P. Sooraj, Prediction of the tensile strength of
IEEE J. Solid State Circ 42 (1) (2007) 56–65. polylactic acid fused deposition models using artificial neural network technique,
[36] M.S. Carvalho, A.P. Martins, T.G. Santos, Simulation and validation of Materials Today: Proceedings, in press 46 (2021) 9187–9193.
thermography inspection for components produced by additive manufacturing, [60] K.P.M.K.V. Dhinakaran, P.M. Bupathi Ram, M. Ravichandran,
Appl. Therm. Eng. 159 (2019), 113872. M. Vinayagamoorthy, A review on recent advancements in fused deposition
[37] M.Y.L.J.H. Park, S.Y. Kwon, H.J. Roh, M.S. Koo, S.H. Cho, Temperature analysis of modeling, Mater. Today:. Proc. 27 (2020) 752–756.
nozzle in a FDM type 3D printer through computer simulation and experiment, [61] S. Vyavahare, S. Teraiya, D. Panghal, S. Kumar, Fused deposition modelling: a
Elastomers Compos 51 (2016) 301–307. review, Rapid Prototyping Journal 26 (1) (2020) 176–201.
[38] D.H. Morton-Jones, Polymer Processing, Chapman and Hall, 1992. [62] D.R.H. Prajapati, A. Jain, Measurement and modeling of filament temperature
[39] Y. Zhang, Y.K. Chou, Three-dimensional finite element analysis simulations of the distribution in the standoff gap between nozzle and bed in polymer-based additive
fused deposition modelling process, Proceedings of the Institution of Mechanical manufacturing, Addit. Manuf. 24 (2018) 224–231.
Engineers, Part B: Journal of Engineering Manufacture 220 (10) (2006) [63] N.R.S. Miftahur Rahman, L.K. Sadhu, Glass Transition of ABS in 3D Printing.
1663–1671. Excerpt from the Proceedings of the 2016 COMSOL Conference in Boston, 2016.
[40] Y. Zhang, K. Chou, A parametric study of part distortions in fused deposition [64] H.M. Ali, Analysis of heat pipe-aided graphene-oxide based nanoparticle-enhanced
modelling using three-dimensional finite element analysis, Proceedings of the phase change material heat sink for passive cooling of electronic components,
Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture. J. Therm. Anal. Calorim. 146 (2021) 277–286.
222 (8) (2008) 959–968. [65] F. Hassan, F. Jamil, A. Hussain, H.M. Ali, M.M. Janjua, S. Khushnood, M. Farhan,
[41] V.S.B. Courter, J. Bi, S. Dev, C.J. Hansen, Finite element simulation of the fused K. Altaf, Z. Said, C. Li, Recent advancements in latent heat phase change materials
deposition modelling process, Proceedings of the NAFEMS World Congress (2017) and their applications for thermal energy storage and buildings: a state of the art
11–14. review, Sustainable Energy Technol. Assess. 49 (2022), 101646.
[42] D.D. Phan, Z.R. Swain, M.E. Mackay, Rheological and heat transfer effects in fused [66] T.D. J.Brady, P.I.Lee, J.-X.Li, Chapter 7 - Polymer Properties and Characterization,
filament fabrication, J. Rheol 62 (5) (2018) 1097–1107. Pharmaceutical Theory and Practice, (20017) 181-223.
[43] X.Z.S.-J. Hsieh, Thermal analysis of fused deposition modeling process using [67] K.R.M. Lori, R. Hilden, Physics of amorphous solids, Physics of amorphous solids
infrared thermography imaging and finite element modeling, Proc. SPIE 10214, 93 (2004) 3–12.
Thermosense: Thermal Infrared Applications XXXIX, (2017) 1021409. [68] C. Schick, Differential scanning calorimetry (DSC) of semicrystalline polymers,
[44] T.N.Y. Zhou, G. Xiong, D. Liu, Temperature Analysis in the Fused Deposition Anal. Bioanal. Chem. 395 (6) (2009) 1589–1611.
Modeling Process, 3rd International Conference on Information Science and [69] M. Ali, R.K. Sari, U. Sajjad, M. Sultan, H.M. Ali, Effect of annealing on
Control Engineering (ICISCE), Beijing, China, 2016. microstructures and mechanical properties of PA-12 lattice structures proceeded by
[45] T.R.D.R.V. Nancharaiah, D.R. Raju, V.R. Raju, An experimental investigation on multi jet fusion technology, Addit. Manuf. 47 (2021), 102285.
surface quality and dimensional accuracy of FDM components, International [70] M.D.D. Riedlbauer, D. Drummer, P. Steinmann, J. Mergheim, Modelling,
Journal on Emerging Technologies 1 (2) (2010) 106–111. simulation and experimental validation of heat transfer in selective laser melting of
[46] K.P.P.M. Thrimurthulu, P.M. Pandey, N.V. Reddy, Optimum part deposition the polymeric material PA12, Comput. Mater. Sci. 93 (2014) 239–248.
orientation in fused deposition modeling, Int. J. Mach. Tools Manuf 44 (6) (2004) [71] R.S. Ranvijay Kumar, I. Farina, On the 3D printing of recycled ABS, PLA and HIPS
585–594. thermoplastics for structural applications, PSU Research Review 2 (2014) 115.
[47] A.H.P.J.W. Zhang, Process-parameter optimization for fused deposition modeling [72] S. Mazloum, S. Awad, N. Allam, Y. Aboumsallem, K. Loubar, M. Tazerout, Khaled
based on Taguchi method, Adv Mater Res 538 (2012) 444–447. Loubar, Mohand Tazerout, Modelling plastic heating and melting in a semi-batch
[48] T.W.L. C. C. Wang, S. S. Hu Optimizing the rapid prototyping process by pyrolysis reactor, Appl. Energy 283 (2021) 116375.
integrating the Taguchi method with the gray relational analysis, Rapid Prototyp J, [73] A. Ejaz, F. Jamil, H.M. Ali, A novel thermal regulation of photovoltaic panels
13 (2007) 304-315. through phase change materials with metallic foam-based system and a concise
[49] G.F.C.H.Y. Wan, C.P. Li, X.B. Qi, G.P. Zhang, Data-driven evaluation of fatigue comparison: An experimental study, Sustainable Energy Technologies and
performance of additive manufactured parts using miniature specimens, J. Mater. Assessments, 49 (2022) 101726.
Sci. Technol. 35 (2019) 1137–1146. [74] S.H. Huang, H.-C. Zhang, Artificial neural networks in manufacturing: concepts,
[50] Q.S. Ruishan Xie, G. Chen, Improved distortion prediction in additive applications, and perspectives, IEEE Trans. Compon. Packag. Manuf. Technol. Part
manufacturing using an experimental-based stress relaxation model, J. Mater. Sci. A: 17 (1994) 212–228.
Technol. 59 (2020) 83–91. [75] W. Wu, W. Deng, Y. Huang, X. Wang, Y. Ji, Prediction of the working conditions for
[51] M.G.Z.N. Moslemi, A. Ayob, N. Redzuan, S. Rhee, Evaluation of sensitivity and the pulse tube cooler based on artificial neural network model, Appl. Therm. Eng.
calibration of the chaboche kinematic hardening model parameters for numerical 197 (2021), 117424.
ratcheting simulation, Applied Sciences 9 (2019) 2578. [76] K. Zhang, Z. Zhang, Y. Han, Y. Gu, Q. Qiu, X. Zhu, Artificial neural network
[52] N. Moslemi, M. Golzardian, N. Redzuan, F. Mozafari, A. Ayob, Optimization modeling for steam ejector design, Appl. Therm. Eng. 204 (2022), 117939.
procedure for parameter determination of caboche kinematic hardening model, [77] J. Munguía, J. Ciurana, C. Riba, Neural-network-based model for build-time
IOP Conference Series: Materials Science and Engineering 884 (1) (2020) 012112. estimation in selective laser sintering, Proceedings of the Institution of Mechanical
[53] F.M. N Moslemi*, B Abdi, S Gohari, N Redzuan*, C Burvill, A Ayob, Uniaxial and Engineers, Part B: Journal of Engineering Manufacture 223 (8) (2009) 995–1003.
biaxial ratcheting behavior of pressurized AISI 316L pipe under cyclic loading: [78] P. Gupta, P. Kumar, S.M. Rao, Artificial neural network model for single-phase real
Experiment and simulation, N Moslemi, F Mozafari, B Abdi, S Gohari, N Redzuan*, gas ejectors, Appl. Therm. Eng. 201 (2022), 117615.
C Burvill, A Ayob, 179 (2020) 105693. [79] E.P. Oliveira, G.d.M. Stieven, E.F. Lins, J.R. Vaz, An inverse approach for the
[54] S. Motahar, M. Jahangiri, Transient heat transfer analysis of a phase change interfacial heat transfer parameters in alloys solidification, Applied thermal
material heat sink using experimental data and artificial neural network, Appl. engineering, 155 (2019) 365-372.
Therm. Eng. 167 (2020), 114817. [80] C.-H. Huang, Y.-T. Wu, An optimum design for a natural convection pin fin array
[55] H. El-Mounayri, J.F. Briceno, M. Gadallah, A new artificial neural network with orientation consideration, Appl. Therm. Eng. 188 (2021), 116633.
approach to modeling ball-end milling, The International Journal of Advanced [81] A. Fguiri, C. Marvillet, M.R. Jeday, Estimation of fouling resistance in a phosphoric
Manufacturing Technology 47 (5) (2010) 527–534. acid/steam heat exchanger using inverse method, Appl. Therm. Eng. 192 (2021),
116935.

17