Materials and Manufacturing Processes

ISSN: 1042-6914 (Print) 1532-2475 (Online) Journal homepage: http://www.tandfonline.com/loi/lmmp20

Experimental Investigations of Process Parameters

Influence on Rheological Behavior and Dynamic
Mechanical Properties of FDM Manufactured Parts

Omar Ahmed Mohamed, Syed Hasan Masood & Jahar Lal Bhowmik

To cite this article: Omar Ahmed Mohamed, Syed Hasan Masood & Jahar Lal Bhowmik (2015):
Experimental Investigations of Process Parameters Influence on Rheological Behavior and
Dynamic Mechanical Properties of FDM Manufactured Parts, Materials and Manufacturing
Processes, DOI: 10.1080/10426914.2015.1127955

To link to this article: http://dx.doi.org/10.1080/10426914.2015.1127955

Accepted author version posted online: 17

Dec 2015.

Submit your article to this journal

Article views: 9

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=lmmp20

Download by: [University of Nebraska, Lincoln] Date: 29 December 2015, At: 02:59
 Experimental Investigations of Process Parameters Influence on Rheological
Behavior and Dynamic Mechanical Properties of FDM Manufactured Parts

Omar Ahmed Mohamed1, Syed Hasan Masood1, Jahar Lal Bhowmik2

1
Department of Mechanical and Product Design Engineering, Swinburne University of
Technology, Hawthorn, Victoria, Australia,
2
Department of Statistics, Data Science and Epidemiology, Swinburne University of
Technology, Hawthorn, Victoria, Australia
Corresponding author to Omar Ahmed Mohamed: ,. E-mail:
Omar.Ahmed.Mohamed@outlook.com.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

Abstract

Fused deposition modelling (FDM) has gained popularity in industry because of its

ability to manufacture complex parts. But when it comes to the manufacturing of

functional products, the advantages of FDM is not so distinct due to the high number of

intervening parameters and complex optimal settings set up. This paper investigates the

influence of process parameters on the rheological and dynamic mechanical properties of

FDM manufactured parts. In this study, an attempt has been made to establish an

empirical relationship between the FDM input parameters and the properties involved

using IV- optimal response surface methodology and statistical analysis. Further,

optimized process parameters were established to maximize the rheological and dynamic

mechanical properties through the graphical optimization. The optimization results show

that the parameters with the most significant effect on the rheological and dynamic

mechanical properties are the layer thickness, the air gap, the road width and the number

of contours. The results also show that by taking into consideration the number of

contours, the functionality of manufactured part is improved significantly.

1
 KEYWORDS: Fused; Parameters; Modulus; Dynamic; Flexural, Viscosity, ANOVA,

Viscoelastic, Mechanical

INTRODUCTION

Fused deposition modelling (FDM) is a widely used and highly efficient additive

manufacturing process for producing complex shapes at low cost without using any tools.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

FDM is a 3D printing technology that builds parts from a from computer aided design

(CAD) data by melting and layering thin coils of material through a heated nozzle (see

Fig. 1) [1]. The cross section of the part is then deposited on the build plate and the

material then cools and another layer is added to the part [2]. The quality of parts

produced by FDM process is a function of material, part geometry, and process

parameters.

Recent years have seen tremendous innovations and rapid developments in commercial

FDM systems in various series such as Idea Series, Design Series, Production Series and

Dental Series. The literature reveals that the mechanical properties of fabricated parts can

be significantly improved by the appropriate adjustment in the process parameters.

Throughout the production process, there is a strict relationship between the input

parameters and the mechanical properties of manufactured part. For example,

inappropriate settings of process conditions will cause defects on the built parts, such as

incomplete filling in the interior part structure (gaps and voids), which will affect the

properties and functionality of the final built part by the FDM process. Therefore, it is

2
 critically important to effectively control the influence of parameters during the printing

process.

Rheological (viscoelastic properties) and dynamic mechanical properties are some of the

most important properties of polymers in the processing and performance of FDM

processed parts in various industries including automotive exterior and interior

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

components, electronics, aerospace and medical hardware. Understanding of the

rheological and dynamic mechanical properties in relation to FDM process parameters is

of practical interest for several reasons: First, it is necessary to examine the dynamic

material response as a function of process parameters over a wide range of temperatures.

This will help in understating the response of the manufactured parts for long-term

applications including the deformation of materials under the applied load. Secondly, it is

necessary to determine the optimum processing conditions to build functional products

with desired shapes and properties in order to avoid premature failure when the

manufactured parts are subjected to cyclic stress, forces and vibration in various

applications. Thirdly, the rheological and dynamic mechanical properties of processed

part are highly sensitive to the variations of FDM process parameter and they control the

flow properties of material during the molten state processing time.

Rheological properties play an important role in the area of quality control in various

industries. The automotive and aerospace industries work with many complex materials

whose viscosity and viscoelasticity are of great significance in order to achieve high

product properties and quality. Particularly, the shear rate and temperature dependence of

3
 viscosity determines the resistance to flow and deformation of the product in order to

understand whether the product is fully filled or not by the FDM process, which is

dependent upon the process parameters selected to manufacture the product. Rheological

properties affect all stages of material development across multiple industries from

material stability to material processing, product performance and quality assurance. The

obvious importance of rheological properties in product design and development means

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

that it is essential for the manufactures and designers to have analytical models to

estimate these properties under a wide range of processing parameters setting.

Development of the regression models aim to better understand the complex relationship

between the building parameters and the overall rheological properties of the

manufactured parts. The ability to gather data on rheological behavior of the material

gives the manufacturers an important clue on the dimensional stability of the product.

Knowledge of rheological properties is valuable in predicting the product performance

for long term applications. This type of information enables additive manufacturing users

and designers to optimize the processing conditions required to create high quality and

reliable products.

A relatively large amount of research [3-17] have been performed to investigate the

impact of process conditions on static mechanical properties (static loading) of the

Acrylonitrile butadiene styrene (ABS) parts fabricated by FDM process using empirical

models and applied design of experiments through statistical techniques such as

regression analysis and analysis of variance (ANOVA). Unfortunately, a systematic

experimental investigation of the effect of input parameters on the rheological and

4
 dynamic mechanical properties of part manufactured by FDM does not exist in the

literature. However, establishing such a rigorous relationship between the fabrication

parameters, rheological and dynamic mechanical properties is not a simple task and it has

not yet been studied due to the complexity of the FDM process as well as due to the

difficulty in undertaking the rheological and dynamic mechanical properties of FDM

manufactured parts. This research work has been undertaken in an effort to address this
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

research gap. The aim of this study is to establish rigorous empirical relationships

between the input parameters and the rheological and dynamic mechanical properties of

the FDM built parts. The influence of layer thickness, air gap, raster angle, build

orientation, road width and the number of contours on the rheological and dynamic

mechanical properties of the manufactured Polycarbonate/Acrylonitrile-Butadiene-

Styrene (PC/ABS) parts was investigated in order to produce functional parts with

improved mechanical characteristics.

Materials And Methods

Rectangular specimens with dimensions 35 × 12.5 × 3.5 mm were prepared according to

ASTM D5418 [18] standard and TA instrument manufacturer recommendations [19].

The CAD models of the testing specimens are modelled in Pro Engineer Wildfire 5.0 and

exported as an STL (STereoLithography) file. The STL file was then imported to the

FDM software (Insight version 9.1) to slice and create the tool path for the parts and to

set all process parameters considered in this study (see Table 1) for all specimens as per

experimental design matrix as shown in Table 2. All test specimens were manufactured

using Stratasys FDM Fortus 400 machine. The specimens are made in PC-ABS material

5
 developed by Stratays, Inc. The Polycarbonate/Acrylonitrile-Butadiene-Styrene (PC-

ABS) is a blend of polycarbonate and ABS plastic and it has amorphous structures of

based polymer.

Rheological properties (viscoelastic properties) and dynamic mechanical properties

(DMA) of the manufactured specimens were obtained by using a Dynamic Mechanical

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

Analyzer (DMA) (model 2980, TA Instrument), where an oscillatory stress is applied to

the specimen and the resulting strain (displacement) is measured. The rheological and

DMA tests were performed on each test sample in a single cantilever mode on a dual

cantilever clamp at a fixed frequency of 1 Hz and the vibrational amplitude of 15 µm.

The entire temperature range from 35 °C to 170 °C was used in DMA test with a heating

rate of 3°C/min and with an isothermal soak time of 5 min. Fig. 2 shows the arrangement

schematically. The measurements for maximum values of dynamic flexural modulus (

DFM ) and complex viscosity (η) were obtained for each experimental run and then

recorded as the response according to experimental design matrix plan. The average

value for each property was obtained from a set of values of tested specimens. Dynamic

flexural modulus is the ratio of stress to strain under cyclic loading (vibratory) conditions.

It measures the overall resistance to deformation under dynamic load. It is an indicator of

viscoelasticity of a material. The complex viscosity measures the ability of a material or

sample to resist flow and deformation during mechanical oscillation as a function of

temperature, frequency, time, or both. Most polymeric materials behave as viscoelastic.

6
 Table 1 shows the FDM six process parameters at different levels considered in this

study.

Out of these, the five process parameters namely, layer thickness (A), air gap (B), raster

angle (C), build orientation (D), road width (E) have already been identified as the most

significant FDM process parameters by several researchers [3-17] in their studies of FDM
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

process optimization and therefore these main factors are considered in this paper to

study their effects on the selected responses. The sixth parameter, the ‘number of

contours’ is also considered in this study along with five main process parameters. This

factor affects the strength of the layer around the raster beads and hence may affect the

mechanical properties, and therefore it is considered of particular interest to include it in

this study. Other FDM parameters are kept constant as they do not have significant

influence on the responses as stated by the previous studies [3-17]. The levels of each

process parameters were selected based on preliminary studies, their significance, real

industrial applications and the permissible low and high levels recommended by the

equipment manufacturer. The process parameters considered in this study are shown

graphically in Fig. 3

In order to investigate the influences of the input parameters involved, the experimental

design was carried out according to response surface methodology based on IV-optimal

design. The IV-optimal design is a method of experimental design generated by a

computer algorithm (also known as I-, Q-, and V-optimality) used in this instance to

serve the purpose of minimizing the integrated prediction variance over the region of

7
 interest [20]. The IV-optimal design is a novel technique recommended to build response

surface designs when the objective is to optimize the factor settings with the process

requiring greater precision in the estimation [21]. The IV-optimal design has

distinguished itself from other experimental design techniques due to the following

reasons:
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

IV-optimal requires less number of runs than classical response surface designs

(i.e. central composite design). For example, a study of six parameters using central

composite design will require 90 runs without considering replications and extra runs to

improve the efficiency of prediction, while IV-optimal requires only 60 runs including

replications and extra model points.

IV-optimal design is a custom design technique, flexible and tends to place the design

points more uniformly than in the classical response surface design such as Central

Composite Design (CCD) and Box-Behnken Design (BBD). Therefore, greater precision

and accuracy in prediction can be achieved through this design.

IV- optimal design supports multifactor constraints to the design space (irregular

experimental region), which is not applicable in standard response surface designs. For

example, excluding a particular area where responses cannot be measured or in cases of

the number of levels are different from one factor to another (e.g. in this study).

IV- optimal design has another advantage over other experimental designs in that

as it can fit higher order polynomial models such as cubic, quartic and fifth order models

or it can fit a custom model.

8
 The IV-optimal design algorithm suggested 38 runs for the six process parameters at

different levels. However, replications and extra unique model points were added in

addition to the 38 runs to reduce the standard error, and to improve the precision of

estimation and to reflect the true behavior of response parameters. The statistical software

packages ‘Minitab version 17 and SAS Visual Statistics’ are used to analyze the

experimental design. Table 2 shows the developed experimental design matrix along with
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

responses.

RESULTS AND DISCUSSION

In this study, the second order polynomial (regression) model that represents the response

surface can be expressed as:

k k k
2
Y 0 i Xi ii X i ij Xi X j (1)
i 1 i 1 i j

where Y denotes the predicted response, X i and X j are the coded input variables, k is

the total number of variables, β 0 is the constant term of the regression equation, i is the

linear regression coefficient, ii is the regression coefficient for quadratic terms, ij is

the regression coefficient for interaction terms, and is the random error.

In the present work, the estimation of the regression coefficients and the ANOVA

technique were conducted through Minitab (version 17). Therefore, following the

calculation of the regression coefficients, the final regression equations for dynamic

flexural modulus ( DFM ) and complex viscosity ( ) in terms of actual values were

obtained as follows:

9
 DFM MPa 1330.30 2143.91* A 3726.82* B 0.41* D 9734.81* E 17.82* F 1149.58* AB 2814
126.50* BF 3669.21* A2 1895.56* B 2 10221.52* E 2
(2)

MPa.sec 203.56 192.17 * A 220.26* B 0.17 * C 0.12* D 1015.79* E 3.28* F 88.84* AB 194.7
99.75* B 2 1.61 10 3 C 2 1.04 10 3 D 2 1011.74* E 2
(3)

ANOVA was employed to estimate and test the regression coefficients in order to

investigate which process parameters have a significant impact on the dynamic

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

properties. The ANOVA results for dynamic flexural modulus and complex viscosity are

presented in Table 3 and 4. Regression terms with the highest partial probability values

(P> 0.1) were eliminated from the regression equations using the backward elimination

method, as it improves the quality of fit the experimental data appropriately. As seen in

Table 3 and 4, the second-order regression model for dynamic flexural modulus and

complex viscosity was significant because the F-statistic (Fisher test) for dynamic

flexural modulus and complex viscosity was 427.32 and 64.5, respectively and the P-

value (probability value) for dynamic flexural modulus and complex viscosity was less

than 0.05 (5% significance level). The P-value of lack-of-fit for dynamic flexural

modulus and complex viscosity were 0.2413 and 0.6492, respectively, indicated that

lack-of-fit is insignificant in relation to the residual error. Insignificant lack of fit is

good, as it implies that the terms left out of model are insignificant. The determination

coefficient measures the goodness-of-fit of the regression models. The higher value of

determination coefficient (close to 1.0) is desirable. As shown in Table 3 and 4, it is clear

that developed relationship for dynamic flexural modulus and complex viscosity have

large values of R 2 , adjusted R 2 , predicted R 2 , which indicates that there is high

correlation between the experimental values and estimated values. Adequate precision is

10
 a measure of single to noise ratio. Value more than 4 is preferable. In this study, adequate

precision for dynamic flexural modulus and complex viscosity is higher than 4, which

implies an adequate signal.

It is important to examine the assumption of ANOVA results through residual analysis

before drawing firmer conclusions. The normal probability plots of the residuals for
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

dynamic flexural modulus and complex viscosity are shown in Fig. 4 (a and b), which

shows that all the residuals are following a straight line, demonstrating that the residual

errors are normally scattered and the regression equations are correctly established. Fig. 4

(c and d) shows the comparisons between predicted and experimental values (actual

values) for the dynamic flexural modulus and complex viscosity. It can be seen that the

predicted values are very much in agreement with the experimental values. This indicates

that the developed models can predict the responses accurately.

Figs. 5 (a-b) show the effect of six process parameters on dynamic flexural modulus and

complex viscosity as the output responses for an optimization design. These figures show

how the dynamic viscoelastic properties change as each factor moves from low level to

higher level with all other factors held constant at their center levels. A steep slope or

curvature of a factor indicates that the dynamic viscoelastic properties are sensitive to

that factor. Figs. 5 (a-b) show that layer thickness has a significant effect on dynamic

flexural modulus and complex viscosity. However, the influence of layer thickness on

complex viscosity is more observed from its curvature. It is clear that with the increase in

layer thickness, there is an increase in dynamic flexural modulus and complex viscosity

11
 up to around center level, and then with the further increase in layer thickness, the

dynamic flexural modulus and complex viscosity also start decreasing. A possible

explanation is that higher layer thickness means fewer layers and fewer thermal cycles

are required, thereby reducing the non-uniform temperature gradients, distortion and

thermal stresses. Thus, stronger part to deformation resistance and dense structure can be

obtained. Results from previous studies [17, 22, 23] indicated that the optimal level of
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

layer thickness is 0.3302 mm, but the findings from this study showed that the optimal

level of layer thickness is around 0.2794 mm (between the center and higher levels).

However, due to the unavailability of FDM tip diameter (size) to provide layer thickness

of 0.2794 mm, the FDM tip size available gives layer thickness 0.3302, which is the

closest optimal value of 0.2794 mm.

Typical micrographs by scanning election microscope (SEM) are included in this study to

show the different defects observed in the manufactured samples. Fig. 6 (a) is a typical

micrograph for sample fabricated with layer thickness of 0.127 mm showing porosity and

pin holes. When the part is manufactured with the lower value of layer thickness (0.127

mm), then more number of layers is needed to build the part, which leads to low quality

of the cross-link between interlayers, low modular diffusion and non-uniform

temperature. This leads to the formation of pin holes, pores and delamination in the

microstructure as shown in Fig. 6 (a). This phenomenon of inhomogeneity in

microstructure is responsible for producing weak parts.

12
 Figs. 5 (a-b) show air gap is one of the most influential factors on dynamic viscoelastic

properties. This is evidenced by its large steep slope or curvature. It can be seen from

Figs. 5 (a-b) that with the increase in air gap, a prominent decrease in dynamic flexural

modulus and complex viscosity is noticed. This is because the lower value of air gap

means that the built layers are closer to each other, resulting in stronger interfacial

bonding strength between internal rasters or across filaments (less porosity) and dense
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

microstructure. If the part is built with positive air gap, then there will be pores and

porosity between the interlayer. This makes the final built part weaker than the part

processed with lower value of air gap. This result is consistent with the published study

[3, 17, 22]. Although positive value of air gap makes the part weaker, it has some

benefits, as it increases the chance of the fluidity of deposited filament to flow and fill up

the porosity in the interlayer compared with the case of no air gap. Furthermore, it

increases the coefficient of heat transfer and thermal conductivity resulting in reducing

the chance in the development of stresses accumulation responsible for inner micro-

cracks.

Figs. 5 (a-b) also show that raster angle has marginal effect on dynamic flexural modulus

and it has more influence on complex viscosity. It can also be seen from this figure that

with the increase in raster angle from low level (0˚) to center level (45˚), a slight decrease

is observed in dynamic flexural modulus, while a prominent decrease has been seen in

complex viscosity. With the further increase in raster angle beyond the center level (45˚),

the dynamic flexural modulus and complex viscosity also starts increasing. It can also be

noticed that the lowest value of raster angle is preferred in improving dynamic

13
 viscoelastic properties. This is due to the fact that the lowest value of raster angle results

in shorter rasters length, smaller curves, and smaller sharp turns, thereby reducing

incomplete filling between rasters and the perimeter walls and this results in dense

structure, and hence strong interior bonding for the prototype. Moreover, it can be seen

from Figs. 5 (a-b) that the highest value of raster angle (90˚) provides better dynamic

viscoelastic properties than when using center level of raster angle (45˚). However, a

raster angle of 90˚ can achieve better dynamic viscoelastic properties than a raster angle
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

of 45˚ but lower than the raster angle of 0˚. It is well-known that the raster angle from

0˚to 45˚ is similar to the raster angle from 45˚to 90˚. However, this study shows that the

raster angle of 0˚ provides better mechanical properties than the raster angle of 90˚. Fig. 6

shows the difference between the specimen processed with raster angle of 0˚ and the

specimen processed with raster angle of 90˚. The advantage of raster angle at 0˚ over the

raster angle at 90˚ is that the raster angle of 0˚ requires less number of rasters. For

example, the specimen used in this study has a thickness of 3.5 mm. If the specimen is

processed with layer thickness of 0.3302 mm, then the total number of layers required is

10.59. Due to the constant flow rate of extruded filament, the machine will deposit 11

layers. In this case, if the specimen is manufactured with raster angle of 0˚, then the

bottom layer will be deposited with raster angle of 0˚ and the next layer will be deposited

with raster angle of 90˚ and so on until the layer 11. Thus, there will be six layers

deposited with the raster angle of 0˚ and five layers deposited with the raster angle of 90˚.

Since the majority of layers are deposited with the raster angle of 0˚, the specimen shows

better mechanical properties. A possible explanation for this phenomenon is that lower

raster angle (0˚) provides less number of rasters, which are prone to less distortion, and

14
 thus provides stronger interlayer bonding. Furthermore, the deposited layer with raster

angle of 0˚ is subjected to progressively increasing bending load, as shown in Fig. 6 (a),

the stress will require separating a greater number of deposited plastic fibers in the built

specimen, which exhibits higher yield strength and good ductility. However, if the

bending load is applied on the layer, which is deposited with raster angle of 90˚ (see

Fig.6. (b)), the filled specimens consistently exhibit a higher level of brittleness, poor
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

ductility and lower strength (separation occurs at the interface between the melted

layers).

Fig. 7 (b-c) is a typical micrograph for samples manufactured with different raster angles

showing relatively large voids and tear in the manufactured part. The voids and tear are

formed due to long raster length and inappropriate process parameter combinations,

which causes sharp turns, resulting in incomplete filling of interior regions in the

fabricated parts.

In the Figs. 5 (a-b), it is seen that build orientation does not show a significant effect. But

the decrease in the build orientation from 90˚ to 0˚ helps although not significantly. This

is due to the fact that if the part is oriented at XY (45˚), then the built layers of curved

surface will be stacked, resulting in stair-stepping imperfection effect, thus causing some

voids and porosity in the built part, resulting in less dense structure. This observation is

consistent with the much of the previous work [3, 17, 22, 23]. Therefore, orienting the

part on exact X-Y building plane helps in obtaining better curve definition for rasters and

15
 reduces the chance of stair-stepping and voids formation, and the interior structure is also

more intense.

Road width shows marginal effect on dynamic flexural modulus, but it is more influential

on complex viscosity as shown in Figs. 5 (a-b). It can be seen that with the dynamic

flexural modulus and complex viscosity are increased with the increase in road width
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

from 0.4572 mm to 0.5177 mm. With the further increase in road with beyond 0.5177

mm, the dynamic flexural modulus and complex viscosity start decreasing. This is

probably because 0.5177 mm provides a finer path to fill interior rasters and thus good

chance to reduce large sharp turns and incomplete filling. Furthermore, 0.5177 mm of

road width provides more uniform heating and cooling of a material that is proportional

to the thermal expansion coefficient of the extruded material. This can reduce thermal

stress, distortion and thermal gradients in interior rasters. Thus, dynamic viscoelastic

properties of filled specimens consistently exhibit a higher level of strength to

deformation and good ductility. These findings are quite different from some of previous

studies [3, 17, 22] that have demonstrated the lower value of road width gives the optimal

condition. This is because current study considers the rheological behavior and dynamic

mechanical properties, which are not studied in the previous studies. Moreover, the

previous studies have used classical response surface designs with only three levels for

each factor while this study considers 4 to 6 levels for each factor using the IV-optimal

design methodology, which provides better accuracy for prediction than the classical

methods. It is well-known in statistical sciences that increasing number of factor levels

result in reflection of true behavior of the response in relation to process parameters.

16
 Number of contours shows a significant effect on dynamic flexural modulus and complex

viscosity. It should be noted that the default machine setting for this factor is a single

contour. However, from Figs. 5 (a-b), it can be observed that with the increase in number

of contours from lower level (1) to the higher level (10), a significant improvement in

dynamic flexural modulus and complex viscosity is observed. It is because with the
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

increase in number of contours, the number of rasters and rasters length are reduced. This

helps to increase the chance to eliminate gaps and porosity between the interlayers and

thus high resistance to deformation is obtained. Furthermore, a higher value in number of

contours can increase the heating time and can effectively melt the inter-layers, resulting

in more dense structures of processed part. In addition, higher number of contours builds

a stronger part perimeter, thus preventing the failure of the part because the load and

stress applied on the part are carried by the contours, rather than rasters. Fig. 7 (d) is a

typical micrograph for sample manufactured with higher number of contours. As

discussed earlier, higher number of contours is preferred in optimizing the dynamic

mechanical properties, and the samples manufactured with higher number of contours are

found to have fully dense structure.

A high interaction was observed between some process parameters for dynamic flexural

modulus and complex viscosity. Fig. 8 represents the interaction effect of 2D response

surface plots between the most significant process parameters. Empirical relationships

developed in this study show an interesting phenomenon in the dynamic viscoelastic

properties through the interaction effect, which can be seen from the response surfaces

17
 graphs plotted between the air gap and the number of contours as presented in Fig. 8 (a-

b). As discussed earlier, the positive air gap reduces the dynamic mechanical properties

significantly. However, the interaction effect plots (see Fig. 8 (a-b)) show that it is

possible to obtain high dynamic mechanical properties using higher number of contours

(10) with positive air gap (0.5 mm). This is due to the fact that the manufactured part

under this process conditions is still solid and functional as it has fewer porosity and
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

voids. This can be confirmed from Fig. 7 (d).Therefore, the benefit of using higher value

of air gap along with 10 contours will reduce fabrication cost by means of reducing

manufacturing time and whilst still maintaining high functionality of built part and the

potential to create products with versatile uses in many industrial applications.

Fig. 8 (c-d) shows the contour response surface plots for interaction effect between air

gap and road width on dynamic flexural modulus and complex viscosity. In both cases,

with the increase in road width from a lower level to a higher level the dynamic flexural

modulus and complex viscosity are decreased. Because, the lower value of road width

and a lower air gap provides a finer and thinner raster width, which helps in reducing the

porosity, resulting in an increase in the part properties.

In Fig. 8 (e-f) it can be seen that the interaction between layer thickness and air gap

shows a significant effect on dynamic flexural modulus and complex viscosity. In both

cases, it can be observed that with the increase in layer thickness from low level (0.127

mm) to higher level (0.3302) there is a continuing increase in both of the mechanical

properties. This occurs because, in case of using thick layers, fewer layers are required

18
 and thick wall for FDM parts is obtained, resulting in dense fill structure with no gap

between the adjacent rasters, and hence a high mechanical performance is obtained.

Graphical optimization was carried out to determine the optimal process conditions that

yield the highest dynamic mechanical properties. Graphical optimization allows us to

create plots called overlay plots, which are extremely practical for quick technical use to
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

optimum settings of process parameters. The graphical optimization plot as shown in Fig.

9 defines a yellow feasible region (sweet spot) which simultaneously meets all the criteria

for responses. Grayed regions on the graphical optimization plot do not meet the selection

criteria. Red regions on graphical optimization plot are outside the interval bound for at

least one response but inside the constraints. The optimization plot of Fig. 9 generated by

Design Expert software clearly shows that, whatever the optimization criteria, the layer

thickness has to be around its center level of 0.2540, while the number of contours has to

be at its highest level (10 contours) to achieve the maximum dynamic mechanical

properties along with other factors as follows: air gap of 0.007 mm, raster angle of

3.681˚, build orientation of 0.00˚ and road width of 0.474 mm. This result supports the

discussion made earlier on the effect of the input parameters on the responses. Once the

optimum level of the process parameters is determined, the final step is to confirm the

improvement in the dynamic viscoelastic properties derived from the graphical

optimization. Therefore, five additional specimens were manufactured to confirm the

experiment and to validate the optimum parameter settings. The average of the results

from the laboratory experiments obtained from optimized parameter settings is compared

19
 with the predicted values derived from the proposed method using mathematical models

as shown in Table 5. The error percentage was calculated by Eq. (4).

Actual value Predicted value

Prediction error % | | 100 (4)
Predicted value

It can be seen from Table 5 that the variation (error) between the experimental values and

predicted values lies within 0.187% and 1.196% % for dynamic flexural modulus and

complex viscosity respectively. Thus the prediction performance of the model is

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

satisfactory. Clearly, this confirms an excellent achievement of results using the IV-

Optimal response surface method.

CONCLUSIONS

This study was aimed to examine the influence of FDM process parameters on the

dynamic viscoelastic properties. The IV- optimal RSM and statistical analysis were used

in this study to establish empirical relationships between process parameters and the

selected property. Based on the results obtained in this study, the following conclusions

can be drawn:

Dynamic flexural modulus and complex viscosity are highly affected by layer

thickness, air gap, road width and number of contours.

The findings of this study indicate that the variation in raster angle and build orientation

have marginal effect on dynamic flexural modulus and complex viscosity, but not very

significantly. However, they can help in maximizing the dynamic viscoelastic properties.

20
 From SEM analysis, it is seen that the microstructure of the FDM processed part

is affected significantly by the FDM process parameters, which is function of layer

thickness, air gap, road width and number of contours. The experiments have shown that

parts fabricated with higher number of contours are strong, solid and isotropic in contrast

with single contour (default machine setting).

Layer thickness of 0.2540 mm, air gap of 0.007 mm, raster angle of 3.681˚, build
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

orienttaion of 0.00˚, road width of 0.474 mm and 10 numbers of countours were

determined as the optimum process parameters to reach the maximum dynamic flexural

modulus and complex viscosity.

Verification of the experiment has shown that a significant relationship exists

between the estimated results and experimental results. This confirms that developed

empirical relationships are promising.

According on the experimental results, this study can be used as a guide for industry and

future researches and the results provide a good reference to practitioners in FDM

process parameters optimization.

REFERNCES

1. Mohamed, OA.; Masood, SH.; Bhowmik, JL. Optimization of fused deposition

modeling process parameters: a review of current research and future prospects.

Advances in Manufacturing 2015, doi:10.1007/s40436-014-0097-7, 3, 42-53.

2. Gebhardt A. Understanding Additive Manufacturing. Munich: Carl Hanser Verlag

GmbH & Co. KG; 2012.

21
 3. Onwubolu, GC.; Rayegani, F. Characterization and Optimization of Mechanical

Properties of ABS Parts Manufactured by the Fused Deposition Modelling Process.

International Journal of Manufacturing Engineering 2014, doi:10.1155/2014/598531, 13.

4. Savvakis, K.; Petousis, M.; Vairis, A.; Vidakis, N.; Bikmeyev, A. Experimental

Determination of the Tensile Strength of Fused Deposition Modeling Parts. In ASME

2014 International Mechanical Engineering Congress and Exposition. American Society

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

of Mechanical Engineers 2014, doi:10.1115/IMECE2014-37553, V014T011A022-

V014T011A022.

5. Panda, BN.; Bahubalendruni, MR.; Biswal, BB. A general regression neural

network approach for the evaluation of compressive strength of FDM prototypes. Neural

Computing and Applications 2014, doi: 10.1007/s00521-014-1788-5, 1129-1136.

6. Panda, BN.; Bahubalendruni, MR.; Biswal, BB. Comparative evaluation of

optimization algorithms at training of genetic programming for tensile strength prediction

of FDM processed part. Procedia Materials Science 2014, 5,2250-2257.

7. Górski, F.; Kuczko, W.; Wichniarek, R.; Hamrol, A. Computation of Mechanical

Properties of Parts Manufactured by Fused Deposition Modeling Using Finite Element

Method. In 10th International Conference on Soft Computing Models in Industrial and

Environmental Applications 2015, 403-413.

8. Khuong, TL.; Gang, Z.; Farid, M.; Yu, R. Izod Impact Strength of Acrylonitrile

Butadiene Styrene (ABS) Matetials after Used in UP2 3D-Printer. In Applied Mechanics

and Materials. Trans Tech Publ; 2015, 2737-2740.

22
 9. Gajdoš, I.; Kaščák, Ľ.; Spišák, E.; Slota, J. Flexural Properties of FDM Prototypes

Made with Honeycomb and Sparse Structure. In Key Engineering Materials. Trans Tech

Publ; 2015, 169-173.

10. Nidagundi, VB.; Keshavamurthy, R.; Prakash, C. Studies on Parametric

Optimization for Fused Deposition Modelling Process. Materials Today: Proceedings

2015, 2,1691-1699.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

11. Lanzotti, A.; Grasso, M.; Staiano, G.; Martorelli, M.; Pei, E.; Campbell, RI. The

impact of process parameters on mechanical properties of parts fabricated in PLA with an

open-source 3-D printer. Rapid Prototyping Journal 2015, Doi: 10.1108/RPJ-09-2014-

0135, 21, 604 - 617.

12. Brensons, I.; Polukoshko, S.; Silins, A.; Mozga, N. FDM Prototype Experimental

Research of Processing Parameter Optimization to Achieve Higher Tensile Stress. In

Solid State Phenomena. Trans Tech Publ; 2015, 767-773.

13. Impens, D.; Urbanic, R.; Assessing the Impact of Post-Processing Variables on

Tensile and Compression Characteristics for 3D Printed Components. IFAC-

PapersOnLine 2015, 48,652-657.

14. Durgun, I.; Ertan, R. Experimental investigation of FDM process for

improvement of mechanical properties and production cost. Rapid Prototyping Journal

2014, 20,228-235.

15. Arivazhagan, A.; Saleem, A.; Masood, S.; Nikzad, M.; Jagadeesh, K. Study of

Dynamic mechanical properties of fused deposition modelling processed ULTEM

material. American Journal of Engineering and Applied Sciences 2014, DOI:

10.3844/ajeassp, 7, 307-315.

23
 16. Gurrala, PK.; Regalla, SP. Multi-objective optimisation of strength and

volumetric shrinkage of FDM parts: A multi-objective optimization scheme is used to

optimize the strength and volumetric shrinkage of FDM parts considering different

process parameters. Virtual and Physical Prototyping 2014, 9,127-138.

17. Sood, AK.; Ohdar, RK.; Mahapatra, SS. Experimental investigation and empirical

modelling of FDM process for compressive strength improvement. Journal of Advanced

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

Research 2012, 3,81-90.

18. ASTMD5418-07. Standard Test Method for Plastics: Dynamic Mechanical

Properties: In Flexure (Dual Cantilever Beam). West Conshohocken: ASTM

International; 2007.

19. DMA2980. Dynamic Mechanical Analysis. Operator’s Manual. New Castle: TA

Instrument; 2002,4-8.

20. Douglas CM. Design and analysis of experiments. John Wiley & Sons, Inc USA

2001.

21. Myers RH, Montgomery DC, Anderson-Cook CM: Response surface

methodology: process and product optimization using designed experiments. John Wiley

& Sons; 2009.

22. Sood, AK.; Ohdar, R.; Mahapatra, S. Parametric appraisal of mechanical property

of fused deposition modelling processed parts. Materials & Design 2010, doi:

doi:10.1016/j.matdes.2009.06.016, 31,287-295.

23. Eujin, Pei D.; Lanzotti ,A.; Grasso, M.; Staiano, G; Martorelli, M. The impact of

process parameters on mechanical properties of parts fabricated in PLA with an open-

source 3-D printer. Rapid Prototyping Journal 2015, 21,604-617.

24
 TABLE 1. Control factors (parameters) and their levels

Symbols Input Units Levels

FDM 1 2 3 4 5 6

factors

A Layer mm 0.1270 0.1778 0.2540 0.3302 - -

thickiness
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

B Air gap mm 0 0.1 0.2 0.3 0.4 0.5

C Raster degree 0 15 30 45 60 90

angle

D Build degree 0 30 45 60 75 90

orientation

E Road mm 0.4572 0.4814 0.5056 0.5298 0.5540 0.5782

width

F Number - 1 3 5 7 8 10

of

contours

25
 TABLE 2. IV-optimal design matrix and responses

S. Coded factors Responses S. Coded factors Responses

N A B C D E F D FM η N A B C D E F DFM η

o. (MPa) (MPa.se o (MPa) (MPa.se

c) c)

1 3 6 6 6 1 1 448.8 29.17 31 4 6 4 3 1 6 1190. 65.82

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

7 97

2 3 4 4 3 4 3 935.2 51.69 32 4 6 6 6 6 6 1201. 65.77

0 97

3 3 4 4 3 4 3 957.8 52.94 33 4 6 1 1 6 6 1186. 65.55

1 16

4 4 5 6 2 2 1 538.7 25.35 34 4 3 5 6 6 1 738.7 40.83

1 32

5 1 3 4 6 1 1 728.4 36.00 35 3 6 1 6 6 4 1044. 57.74

0 68

6 3 3 4 6 1 3 969.2 53.57 36 4 3 1 1 6 1 832.2 49.15

0 77

7 1 1 1 3 1 3 1293. 71.48 37 4 1 1 4 6 3 1305. 72.17

31 81

8 4 6 1 6 2 1 429.0 23.71 38 3 4 4 3 4 3 961.4 53.14

1 67

9 4 1 6 6 1 5 1398. 80.89 39 1 6 1 6 6 1 516.8 38.68

57 81

26
 10 1 1 4 1 3 4 1343. 74.24 40 3 4 4 3 4 3 922.0 50.96

34 77

11 4 6 6 1 6 1 504.2 27.87 41 3 1 4 6 3 4 1374. 75.96

8 38

12 4 3 1 6 1 6 1221. 74.42 42 1 1 4 1 3 4 1268. 70.12

54 85
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

13 4 4 6 1 1 4 980.5 54.19 43 2 6 4 2 6 1 533.0 29.46

2 29

14 2 6 1 1 1 1 468.5 25.89 44 1 3 4 3 6 6 1144. 63.27

2 8

15 1 1 6 3 1 6 1326. 69.81 45 1 1 1 1 6 1 1096. 60.62

10 83

16 1 6 6 5 3 3 769.2 42.52 46 3 1 1 1 1 6 1460. 80.70

7 16

17 1 6 6 1 6 6 1262. 69.80 47 1 4 6 1 1 1 716.8 39.62

92 3

18 4 1 6 4 2 1 1301. 71.90 48 3 4 6 2 3 6 1235. 71.53

00 26

19 1 3 4 3 6 6 1175. 47.33 49 2 1 1 6 6 6 1289. 71.26

57 46

20 1 6 3 5 5 5 1014. 56.07 50 2 4 6 6 4 6 1158. 61.55

48 56

21 4 1 3 1 1 1 1311. 72.48 51 3 2 6 2 6 3 1052. 58.14

27
 37 02

22 3 6 6 6 1 6 1190. 65.78 52 3 4 4 3 4 3 941.7 52.05

28 22

23 1 1 6 6 6 2 1096. 60.60 53 4 3 5 6 6 1 737.0 40.73

43 18

24 4 1 3 3 4 6 1420. 67.33 54 4 1 6 1 6 6 1367. 75.59

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

19 69

25 2 1 6 1 4 1 1251. 75.73 55 3 4 1 3 1 3 887.2 49.04

24 6

26 3 6 6 5 5 1 561.5 31.04 56 1 4 1 3 3 6 1180. 65.25

6 6

27 2 6 5 1 1 6 1171. 64.76 57 4 6 4 1 3 3 829.3 45.84

79 79

28 2 6 1 6 3 1 1249. 73.43 58 1 5 2 4 3 2 677.8 37.46

57 92

29 3 4 4 3 4 3 972.1 53.72 59 1 6 2 6 1 6 1171. 64.75

0 54

30 1 4 1 1 4 4 1057. 63.51 60 3 4 4 3 4 3 911.5 50.38

05 7

28
 TABLE 3. ANOVA results for dynamic flexural modulus after eliminating insignificant

terms ( P > 0.1)

Source Sum of square Mean sum F P Remarks At

of square 95% CI

Model 4696539 426958 427.32 < Significant

0.0001
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

Linear 3733841 746768 747.40 < Significant

0.0001

A 12518 12518 12.53 0.001 Significant

B 1786081 1786081 1787.60 < Significant

0.0001

D 12293 12293 12.30 0.001 Significant

E 2854 2854 2.86 0.097 Not

significant

F 1838963 1838963 1840.52 < Significant

0.0001

Interaction 562334 187445 187.60 < Significant

0.0001

AB 21851 21851 21.87 < Significant

0.0001

BE 45400 45400 45.44 < Significant

0.0001

BF 518047 518047 518.49 < Significant

29
 0.0001

Quadratic 174275 1838963 1840.52 < Significant

0.0001

A2 14299 58092 14.31 < Significant

0.0001

B2 164464 14299 164.60 < Significant

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

0.0001

E2 15036 164464 15.05 < Significant

0.0001

Residual 47959 999 - - -

Lack-of-fit 41815 1072 1.57 0.241 Not

significant

Pure error 6144 683 - - -

R 2 = 98.99%, Adjusted R 2 = 98.76%, Predicted R 2 = 98.44% Adequate Precision =

73.053

30
 TABLE 4. ANOVA results for complex viscosity after eliminating insignificant terms (P

> 0.1)

Source Sum of square Mean sum F P Remarks At

of square 95% CI

Model 13829.5 921.97 64.50 < Significant

0.0001
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

Linear 10248.5 1708.09 119.49 < Significant

0.0001

A 43.9 43.87 3.07 0.087 Not significant

B 5134.2 5134.17 359.15 < Significant

0.0001

C 38.6 38.55 2.70 0.108 Not significant

D 39.5 39.51 2.76 0.104 Not significant

E 17.8 17.84 1.25 0.270 Not significant

F 4704.2 4704.15 329.07 < Significant

0.0001

Interaction 2187.7 546.92 38.26 < Significant

0.0001

AB 129.8 129.81 9.08 0.004 Significant

BE 217.1 217.09 15.19 < Significant

0.0001

BF 1863.0 1862.98 130.32 < Significant

0.0001

31
 EF 55.4 55.41 3.88 0.055 Not significant

Quadratic 932.3 186.46 13.04 < Significant

0.0001

A2 127.5 127.53 8.92 0.005 Significant

B2 403.3 403.33 28.21 < Significant

0.0001
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

C2 115.9 115.93 8.11 0.007 Significant

D2 43.0 43.03 3.01 0.090 Not significant

E2 130.1 130.15 9.10 0.004 Significant

Residual 629.0 14.30 - - -

Lack-of-fit 484.6 13.85 0.86 0.649 Not significant

Pure error 144.4 16.04 - - -

R2 = 95.65%, Adjusted R2 = 94.17%, Predicted R2 = 91.26%, Adequate Precision =

31.048

32
 TABLE 5. The comparison between experimental and predicted values for optimum

parameter settings

Response Optimum parameter settings Predicte Actual Error

A B C D E F d values values (%)

Dynamic 0.25 0.00 3.68 0.0 0.47 10 1463.44 1466.1 0.187

flexural 4 7 1 0 4 7
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

modulus (MPa)

Complex 84.1998 85.207 1.196

viscosity

(MPa.sec)

33
 Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

34
Figure 1. Schematic diagram of the FDM process.
 Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

35
Figure 2. Method of clamping in single cantilever DMA device.
 Figure 3. (a) layer thickness, (b) FDM tool path parameters, and (c) build orientations.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

36
 Figure 4. Normal probability plots of regression models for (a) dynamic flexural
modulus, and (b) complex viscosity. Predicted versus actual plots for (c) dynamic
flexural modulus, and (d) complex viscosity.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

37
 Figure 5. Effect of process parameters on (a) dynamic flexural modulus, and (b) on
complex viscosity.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

38
 Figure 6. Failure on the specimen processed with (a) raster angle of 0˚, and (b) raster

angle of 90˚.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

39
 Figure 7. SEM images showing (a) pin holes in the part manufactured with layer
thickness of 0.127 mm, (b) large voids and tear in processed part with raster angle of 45˚,
(c) incomplete filling because of sharp turns in the part manufactured with raster angle of
˚, and (d) part manufactured with 10 contours showing dense structure.
Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

40
 Figure 8. 2D contour plots showing the effect of parameters and their interaction

parameters on dynamic flexural modulus and complex viscosity

Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

41
 Downloaded by [University of Nebraska, Lincoln] at 02:59 29 December 2015

42
Figure 9. Overlay plot shows the region of optimal process parameters.