ASM International 1059-9495/$19.00

JMEPEG DOI: 10.1007/s11665-016-2157-6

Effect of Process Parameters on Dynamic Mechanical Performance of FDM PC/ABS Printed Parts Through Design of Experiment Omar Ahmed Mohamed, Syed Hasan Masood, Jahar Lal Bhowmik, Mostafa Nikzad, and Jalal Azadmanjiri (Submitted November 25, 2015; in revised form April 13, 2016) In fused deposition modeling (FDM) additive manufacturing process, it is often difficult to determine the actual levels of process parameters in order to achieve the best dynamic mechanical properties of FDM manufactured part. This is mainly due to the large number of FDM parameters and a high degree of interaction between the parameters affecting such properties. This requires a large number of experiments to be determined. This paper presents a study on the influence of six FDM process parameters (layer thickness, air gap, raster angle, build orientation, road width, and number of contours) on the dynamic mechanical properties of the FDM manufactured parts using the fraction factorial design. The most influential parameters were statistically obtained through the analysis of variance (ANOVA) technique, and the results indicate that the layer thickness, the air gap, and the number of contours have the largest impact on dynamic mechanical properties. The optimal parameters for maximum dynamic mechanical properties were found to be layer thickness of 0.3302 mm, air gap of 0.00 mm, raster angle of 0.0°, build orientation of 0.0°, road width of 0.4572 mm, and 10 contours. Finally, a confirmation experiment was performed using optimized levels of process parameters, which showed good fit with the estimated values. Keywords

analysis of variance (ANOVA), dynamic mechanical properties, fraction factorial design, fused deposition modeling (FDM), process parameters

1. Introduction Fused deposition modeling (FDM) is the most commonly used 3D printing technology for plastic parts in the modern manufacturing world due to its ability of producing complex geometrical shapes without tooling safely in an office-friendly environment. This process produces three-dimensional parts layer-by-layer directly from computer aided design (CAD) model (Ref 1). The process usually begins with taking the feedstock material in the form of a filament and heating it to the semi-molten state. The semi-molten filament is then extruded through the nozzle, which moves over the build table in three axes motion creating a cross section of three-dimensional object onto the platform (Ref 2). The deposited material cools, hardens, and bonds to the layer beneath it. This process is repeated up to the last layer (Ref 1). Although FDM process is an efficient and reliable process, it has received little attention in industrial applications because producing components with high accuracy and desired properties greatly depend on the FDM process parameters fixed at Omar Ahmed Mohamed, Syed Hasan Masood, Mostafa Nikzad, and Jalal Azadmanjiri, Department of Mechanical and Product Design Engineering, Swinburne University of Technology, Hawthorn, VIC 3122, Australia; and Jahar Lal Bhowmik, Department of Statistics, Data Science and Epidemiology, Swinburne University of Technology, Hawthorn, VIC 3122, Australia. Contact e-mail: [email protected].

Journal of Materials Engineering and Performance

the time of manufacturing. Since FDM process depends on the selection of several process parameters, the mechanical properties of the final built part by this process are lower than the parts manufactured by conventional processes (e.g., injection molding). The parts manufactured by FDM exhibit the microstructural anisotropy caused by the layer-by-layer nature of the building process (Ref 3). There are two ways to overcome this limitation. One way is by developing new materials with high mechanical performance. Another way is by understanding and optimizing the process parameters so that appropriate adjustments of these parameters can be obtained, which will help in producing parts with more dense structure and hence improved mechanical properties. In the literature, various studies have been published to improve the mechanical properties under static loading conditions through appropriate adjustments in the process parameters, but the progress has been slow because of the complex nature of the FDM process and conflicting influence of these process parameters. For example, Sood et al. (Ref 4) investigated the mechanical properties (i.e., compressive strength) of FDM printed parts under various process parameters, which is the combination of layer thickness, air gap, raster width, raster angle, and build orientation. Results revealed that variations in process parameters have strong impact on compressive strength. Onwubolu and Rayegani (Ref 5) analyzed the effect of build orientation, raster angle, raster width, and air gap on tensile strength of the processed parts. Results indicated that the partÕs strength improved with decreasing air gap and build orientation and increasing raster angle. Percoco et al. (Ref 6) examined the effect of the chemical treatment on the part strength of FDM fabricated parts. Masood et al. (Ref 7) conducted an experimental work to identify the critical FDM operating conditions to produce parts with highest mechanical performance. Recently, Lanzotti et al. (Ref 8) investigated the effect of process parameters (i.e., layer thickness, infill

orientation, and the number of shell perimeters) on the ultimate tensile strength of FDM Polylactic Acid printed parts, and suggested the optimal parameter settings to optimize the ultimate tensile strength. Durgun and Ertan (Ref 9) described how the parts built by FDM with different part orientations and raster angles influence the surface roughness, tensile strength and flexural strength. This study demonstrated that the part orientation has a strong effect on the surface roughness and mechanical performance than the raster angle. Go´rski et al. (Ref 10) developed a method to compute deformation and stresses in the built parts by FDM by considering critical processing variables. Zhang and Chou (Ref 11) developed a finite element model to analyze the distortions and stress distribution at different processing conditions through central composite design (CCD) and analysis of variance ANOVA technique. It was concluded that layer thickness was the main factor affecting the residual stress and distortion in FDM manufactured parts. Impens et al. (Ref 12) examined the influence of post-processing conditions on tensile and compressive strengths for 3D printed prototypes. Arivazhagan and Masood (Ref 13) performed dynamic mechanical tests in order to study the effect of three processing parameters (i.e., built style, raster width, and raster angle) on the dynamic mechanical properties of FDM fabricated polycarbonate parts. Arivazhagan et al. (Ref 14) conducted similar investigation but using ABS material, and through trial and error experiments. However, this study has some limitations because not all critical process parameters were considered, and the trial and error method can only optimize one parameter setting at a time. In the present literature, there has been no systematic study done so far for the optimization of mechanical properties of FDM manufactured part under dynamic and cyclic loading conditions. The FDM manufactured parts are often being used under repetitive and cyclic loading conditions, which may include millions of cycles and their behavior must be properly understood using a systematic and reliable experimental approach. Unlike previous works, this study investigates the influence of six FDM process parameters (layer thickness, air gap, raster angle, build orientation, road width, and number of contours) on the dynamic mechanical properties of FDM processed PC-ABS part. In this study, empirical relationships between the process parameters and dynamic behavior (storage modulus, loss modulus, and mechanical damping) were studied using fractional factorial design and analysis of variance (ANOVA) technique. This study aims to assist in the selection of optimum process parameters involving cyclic loading as a function of temperature.

2. Experimental Procedure For experiments, all test specimens are fabricated using Stratasys FDM Fortus 400 mc. The test specimens are solid rectangular pieces with dimensions 35 9 12.5 9 3.5 mm. The model material used for test specimens is Polycarbonate/ Acrylonitrile-Butadiene-Styrene (PC-ABS) blend, working in conjunction with a soluble support material. PC-ABS blend is a relatively new material for FDM technology. It is an ideal material for the production of prototypes and tooling. The PC gives superior mechanical properties and heat resistance, while the ABS gives flexibility and chemical resistance. PC-ABS blends are widely used in automotive, electronics, and telecommunications applications. This material was selected

Start

Select process parameters and their levels

Develop regression models

Design Plan Two level L16 Orthogonal Fractional Factorial Design

Analyze data & Interpret results

Perform experiment according to design plan & collect data

Optimization

Dynamic mechanical measurement & result Table preparation

Verification

Conclusion

End

Fig. 1 Flowchart of experimental procedure

in this study due to its wide application in modern industry and also to explore more knowledge on its properties, behavior, and performance under the various process conditions. The designed experimental procedure is shown in Fig. 1. In this study, the six FDM process parameters were included for investigation, and shown in Table 1, which are layer thickness (A), air gap (B), raster angle (C), build orientation (D), and road width (E) (Ref 1, 15). This study also included another critical FDM parameter—the number of contours (F), which has not been studied previously for different quality characteristics. Other FDM parameters are kept at their fixed levels. The levels of these control parameters are selected based on the literature review (Ref 4, 5, 8, 9, 16), experience, their significance, relevance according to the preliminary pilot investigations, and the permissible low and high levels recommended by the equipment manufacturer. The process parameters considered in this study are also shown graphically in Fig. 2. To investigate the influence of processing parameters on the dynamic mechanical properties of the FDM manufactured part, the storage modulus, loss modulus, and mechanical damping were set as the response criteria. Dynamic mechanical analysis (DMA) is a proven testing technique for the characterization of the dynamic mechanical performance, where the specimen is subject to a small deformation in a cyclic manner. The storage modulus describes the elastic properties under dynamic condition. It measures the ability of a material or specimen to store or absorb energy (deformation resistance) (Ref 17). Loss modulus measures the viscous properties of the material. Mechanical damping (also known as tan delta) measures the energy dissipation of a material or sample under dynamic and cyclic loading conditions (Ref 18). DMA machine works under the concept of applying a dynamic load to a material and analyzing the material behavior in response to that load. The material is heated at a constant rate of temperature and deformed (oscillated) at a constant amplitude (strain). The correlation between stress and strain is expressed mathematically as the modulus (Ref 19). When the stress is released, the material tries to

Journal of Materials Engineering and Performance

Table 1 FDM process parameters and their levels Levels Factors Controllable factors Layer thickness Air gap Raster angle Build orientation Road width Number of contours

Units

Symbols

mm mm deg deg mm …

A B C D E F

Low

0.127 0 0 0 0.4572 1

High

0.3302 0.5 90 90 0.5782 10

Fig. 2 (a) Layer thickness, (b) FDM tool path parameters, and (c) build orientations. (Sample dimensions: L = 35 mm, W = 12.5 mm, and T = 3.5 mm)

degree. From the produced data points and sample dimensions, the standard data analysis software called Thermal Advantage Universal Software calculates the corresponding stress and strain, then the software uses the corresponding stress and strain to determine the storage modulus G*, loss modulus G**, and mechanical damping (loss angle d) versus temperature. The average maximum value for each mechanical property was determined from a set of values obtained from tested samples for each experimental run according to a fractional factorial design matrix. Figure 3 shows the geometric illustration of dynamic mechanical analysis curve. The storage modulus G*, loss modulus G**, and mechanical damping (loss angle d) are expressed mathematically as follows: G ¼ Fig. 3

Geometric illustration of dynamic mechanical analysis curve

recover to its original position. Once the stress drops down to zero, the deformation occurs on the sample, but it is not completely recoverable to its original state. From analyzing this response, the DMA machine can determine different properties from the recorded dynamic analysis. A huge number of data points can be produced with units made up of smallest increments of less than a

Journal of Materials Engineering and Performance

r0 cos d e0

G ¼

ðEq 1Þ

r0 sin d; e0

ðEq 2Þ

G ; G

ðEq 3Þ

d ¼ ar tan

where r is the calculated stress and e is the calculated strain. DMA measurements were carried out on the PC-ABS fabricated parts using a dynamic mechanical analyzer (model

2980), at a fixed frequency of 1 Hz in a temperature ramp mode. The specimens were then heated at a rate of 3 C/min from temperature range 35–170 C with an isothermal soak time of 5 min. Figure 4 shows the arrangement schematically where the specimen is clamped in a single cantilever mode subjected to cyclic stress. The measurements for maximum values of storage modulus, loss modulus, and mechanical damping were obtained for each experimental run according to the experimental design matrix as shown in Table 2 and then recorded as the response. The average value for storage modulus, loss modulus, and mechanical damping was obtained from a set of values obtained from tested samples. Factorial experimental designs are mainly used to estimate the significant factors and their interaction, but can also be used to model and optimize the process. Factorial designs are classified as full factorial design and fractional factorial design. The main advantage of using full factorial design is its ability to estimate all the main effects of factors and their interactions (Ref 20). However, when the number of factors increases, the use of full factorial design requires an excessive number of experiments (Ref 21). Practically, it is not required to estimate all the interaction effects because the interactions of more than two factors (order ‡ 2) are usually negligible or not important.

Thus, only a certain number of runs (fractional factorial design) can be used instead of using all runs specified by full factorial design. Fractional factorial design sacrifices interaction affects so that the main effects can still be estimated correctly with fewer runs compared to the full factorial design. The fractional factorial design is also the most commonly used type of techniques for process development and quality improvement which is more efficient with respect to time and cost. The precision and efficiency of an experimental design depend on the careful planning and execution of the experimental procedures as well as the accurate measurement of the responses. In order to investigate the effects of the parameters involved, the experimental design was carried out according to a one-quarter fraction of the 2k design which is called a 2k2 fractional factorial design (resolution IV), where k is the number of factors. In this study, the predicted response Y is proposed using a two-factor interaction (2FI) model expressed by the regression equation:

Y ¼ b0 þ

6 X i¼1

bi Xi þ

6 X

bij Xi Xj þ e;

ðEq 4Þ

i
where b0 is the constant term or intercept of the regression equation, bi is the linear regression coefficient, bij is the regression coefficient for interaction terms, Xi and Xj are the coded independent variables, and eis the random error, which contains measurement error and other unknown variability. Considering the six process parameters denoted by A, B, C, D, E, and F as shown in Table 1, the regression model (Eq 4) can be expressed as follows: Y ¼ b0 þ b1 A þ b2 B þ b3 C þ b4 D þ b5 E þ b6 F þ b12 AB þ b13 AC þ b14 AD þ b15 AE þ b16 AF þ b23 BC þ b24 BD þ25 BE þ b26 BF þ b34 CD þ b35 CE þ b36 CF þ b45 DE þ b46 DF þ b56 EF þ e;

Fig. 4

Method of clamping in single cantilever DMA device

ðEq 5Þ

where Y denotes the predicted response, b0 is the constant term, the letters, A, B, C, D, E, F represent the main effects

Table 2 Experimental design matrix Run

A

B

C

D

E

F

Storage modulus, Mpa

Loss modulus, Mpa

Mechanical damping

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

+1 +1 1 1 1 +1 +1 1 1 +1 +1 +1 1 1 +1 1

1 +1 +1 +1 1 +1 1 1 1 1 +1 +1 +1 +1 1 1

+1 1 1 1 1 1 1 +1 +1 +1 +1 +1 +1 +1 1 1

+1 1 +1 1 1 +1 +1 1 +1 1 +1 1 1 +1 1 +1

1 1 +1 +1 1 1 +1 +1 +1 1 +1 +1 1 1 +1 1

1 +1 1 +1 1 1 +1 +1 1 +1 +1 1 1 +1 1 +1

1321.29 1206.51 580.82 1343.00 1258.57 444.33 1306.87 1271.55 1120.49 1472.35 1179.12 495.11 445.40 1129.42 1233.98 1331.46

165.618 112.871 62.920 143.086 144.804 47.063 139.704 135.726 116.495 170.493 117.921 49.759 43.836 120.615 151.773 141.351

0.95417 1.09229 0.85219 0.99898 0.95251 0.72875 1.09590 0.95976 0.96159 1.13223 1.09240 0.81104 0.79521 0.96210 1.00002 0.93393

Journal of Materials Engineering and Performance

Table 3 Correlation matrix of the factors and their interactions

A B C D E F AB AC AD AE AF BD BF

A 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

B 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

C 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

D 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

E 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

F 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

of the factors included in the model, combinations of factors (such as AB) represent an interaction effect between the individual factors in that term, and e is the random error. Note that the regression is done on the basis of the coded units of the variables and the coefficients are based on that coding. The actual equation that is reported is derived from the coded equation. Low and high level of each of the factor is coded into +1 and 1, respectively, using Eq 6. Table 2 shows the fractional factorial design matrix along with the measured responses used in this study.

Xcoded ¼

XActual  X : ðXHigh  XLow Þ=2

ðEq 6Þ

The correlation matrix (see Table 3) shows the correlations between the columns of the design matrix. A perfectly orthogonal design having a diagonal matrix with 1.0000 indicates that the estimates of the coefficients of the effects will be perfectly correlated. In this study, there is no correlation observed between the effects. This means that independent error-free estimates can be done for all of the factors.

3. Results and Discussion Minitab software version 17 was used to analyze the data collected from experiment. The developed regression models in terms of actual values for storage modulus, loss modulus, and mechanical damping including the input variables and their interaction effects are given below in Eq 7-9. StoragemodulusðMPaÞ ¼ 838:87  1929:66  A  1324:27  B  2:95  C þ 803:24  E þ 12:442  F  1294:27AB þ 11:25  AC  3867:97AE þ 135:81  BF ðEq 7Þ

Journal of Materials Engineering and Performance

AB 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

AC 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000

AD 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000 0.0000

AE 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000 0.0000

AF 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000 0.0000

BD 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000

BF 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000

Loss modulus ðMPaÞ ¼ 138:77 þ 31:10  A  128:96  B  0:432  C þ 0:238  F  324:94  AB þ 1:748  AC þ 15:685  BF ðEq 8Þ Mechanical damping ¼ 0:848 þ 0:0125  A  0:234  B  3:926  104  C þ 3:005  104  D þ 0:228  E  0:0125  F  0:63615  AB þ 1:802  103  AC  2:293  103  AD þ 0:0855  AF þ 0:0392  BF ðEq 9Þ Equations 7-9 represent the regression models for the storage modulus, loss modulus, and mechanical damping, and these models contain four linear terms (A, B, C, and F) and three interaction effects (AC, AF, and BF). The positive sign of an effect means the term has a synergistic influence, while the negative sign means the term has a reverse or antagonistic effect. Equations 7-9 show that the air gap (B) has the highest effect on all dynamic mechanical responses. The equations also show that interaction between air gap (B) and number of contours (F) has the strongest effect on storage modulus and loss modulus, while interaction between layer thickness (A) and number of contours has the largest impact on mechanical damping, and it has the largest regression coefficient among the interactions. To assess the significance of the process parameters and their interaction effects, analysis of variance (ANOVA) technique was used. The model terms with larger F value (Fisher F test) and low probability value (P < 0.0500) have a significant effect on the dynamic mechanical responses. Insignificant terms with the highest partial probability values (P > 0.1) have been removed from the model using backward elimination method to make the regression models and empirical relationships more

efficient. Figure 5 (a)-(c) shows Pareto Chart of the standardized effects of process parameters on the responses. According to the results obtained from ANOVA presented in Table 4, the model F values of 90.3, 27.05, and 33.82 for storage modulus, loss modulus, and mechanical damping, respectively, imply that the models are significant. Furthermore, ANOVA results show that air gap and number of contours have a significant effect on all dynamic mechanical responses. However, layer thickness has no significant effect on storage modulus and loss modulus, but it has a significant effect on mechanical damping. Moreover, ANOVA results show that raster angle does not have significant influence on the dynamic mechanical properties. These results also indicate that build orientation and road width do not have significant effect on storage modulus and loss modulus. However, build orientation and road width have significant effect on mechanical damping. Therefore, build orientation was eliminated from the regression models those are created for storage modulus and loss modulus. However, road width was not removed from the regression model created for storage modulus because its interaction effect with layer thickness was found to be significant, but road width was eliminated from loss modulus regression models as it does not have significant effect either individually or in interaction with other model terms. The coefficient of determination (R2) for storage modulus, loss modulus, and mechanical damping were, respectively, 0.9837, 0.9475 and 0.9422, which is close to 1. This means that 98.37, 94.75, and 94.22% of total variation in storage modulus, loss modulus, and mechanical damping can be explained by the empirical models developed in this study, which represent the relationship between experimental and estimated results. The adjusted R2 and the predicted R2 values are in reasonable agreement with the R2 , indicating that there is a strong correlation between the experimental results and predicted results. Adequate precision measures the signal-to-noise ratio and a ratio of higher than 4 is desirable. The signal-to-noise ratio for all of the fitted regression models are greater than 4, which indicates an adequate signal.

3.1 Effect of Layer Thickness

Fig. 5 chart

Screening of important process parameters using a Pareto

Figure 6(a)-(c) shows the main effect plots of different process parameters on dynamic mechanical responses. From this figure, it can be seen that storage modulus, loss modulus, and mechanical damping increase as the layer thickness increases. However, the effect of layer thickness is much more evident on mechanical damping than on other responses. As the layer thickness increases, there are fewer number of layers needed to manufacture the part, resulting in minimum distortion and thermal cycles and thus the dynamic mechanical properties of the built part are improved. Distortion is mainly responsible for the poor quality of the bonded layers, which occurs due to high non-uniform temperature gradient between the bottom and the top layers during the filling of interior rasters. Figure 7(a)(b) shows the scanning electron microscopy (SEM) micrographs of the specimen manufactured with the lower value of layer thickness 0.127 and 0.1778 mm. It can be seen from Fig. 7(a) that there were pinholes in the specimen processed with layer thickness of 0.1778 mm. However, if the specimen is processed with the lowest value of layer thickness of 0.127 mm, the SEM image (see Fig.7(b)) shows the evidence of tearing, perforation, and delamination in the rasters due to

Journal of Materials Engineering and Performance

Table 4 ANOVA results of regression models for different responses Source

Sum of squares

Storage modulus(a) Model 1.909 9 106 A 1999.21 B 7.625 9 105 C 4583.63 E 384.06 F 6.973 9 105 AB 17291.59 AC 42350.55 AE 9044.49 BF 3.735 9 105 Residual 6110.16 Cor total 1.915 9 106 Loss modulus(b) Model 26550.46 A 134.38 B 13682.74 C 33.38 F 5606.23 AB 1089.91 AC 1021.91 BF 4981.92 Residual 513.96 Cor total 27064.43 Mechanical damping(c) Model 0.20 A 0.015 B 0.027 C 1.213 9 105 D 1.620 9 103 E 3.044 9 103 F 0.092 AB 4.177 9 103 AC 1.086 9 103 AD 1.758 9 103 AF 0.024 BF 0.031 Residual 3.396 9 104 Cor total 0.20

DOF

Mean square

F value

Prob > F

Remarks

9 1 1 1 1 1 1 1 1 1 6 15

2.121 9 105 1999.21 7.625 9 105 4583.63 384.06 6.973 9 105 17291.59 42350.55 9044.49 3.735 9 105 1018.36 …

208.28 1.96 748.75 4.50 0.38 684.77 16.98 41.59 8.88 366.74 … …

<0.0001 0.2107 <0.0001 0.0781 0.5617 <0.0001 0.0062 0.0007 0.0246 <0.0001 … …

Significant Not significant Significant Not significant Not significant Significant Significant Significant Significant Significant … …

7 1 1 1 1 1 1 1 8 15

3792.92 134.38 13682.74 33.38 5606.23 1089.91 1021.91 4981.92 64.25 …

59.04 2.09 212.98 0.52 87.26 16.96 15.91 77.54 … …

<0.0001 0.1861 <0.0001 0.4916 <0.0001 0.0033 0.0040 <0.0001 … …

Significant Not significant Significant Not significant Significant Significant Significant Significant … …

11 1 1 1 1 1 1 1 1 1 1 1 4 15

0.018 0.015 0.027 1.213 9 105 1.620 9 103 3.044 9 103 0.092 4.177 9 103 1.086 9 103 1.758 9 103 0.024 0.031 8.489 9 105 …

215.33 177.15 317.93 0.14 19.09 35.86 1081.66 49.21 12.79 20.71 288.11 365.97 … …

<0.0001 0.0002 <0.0001 0.7247 0.0120 0.0039 <0.0001 0.0022 0.0232 0.0104 <0.0001 <0.0001 … …

Significant Significant Significant Not significant Significant Significant Significant Significant Significant Significant Significant Significant … …

(a)R2 = 98.10%, Adjusted R2 = 96.44%, Predicted R2 = 92.40% Adequate Precision = 22.148 (b)R2 = 94.75%, Adjusted R2 = 91.24%, Predicted R2 = 83.40% Adequate Precision =14.998 (c)R2 = 99.83%, Adjusted R2 = 99.37%, Predicted R2 = 97.30% Adequate Precision =50.566

large number of thin layers required to build the part. Thus, dynamic mechanical performances are decreased.

3.2 Effect of Air Gap Air gap has a significant influence on storage modulus, loss modulus, and mechanical damping as shown in Fig. 6(a)-(c). It is observed that with the decrease of air gap, all three mechanical performance properties have significantly increased. This result is expected because no air gap means the rasters are close to each other with no porosity, resulting in a dense structure and stronger adhesive strength between the interlayers or across filaments, and hence it improves the dynamic mechanical properties but gives poor dimensional accuracy. However, if the part is manufactured with higher value of air gap, then the rasters will be deposited far from each other, resulting in low strength and low density of the manufactured parts. In practical applications, the lower value

Journal of Materials Engineering and Performance

of air gap increases the chance of developing stress accumulation by restricting heat loss or heat transfer. In such situations, the part needs longer exposure and more time to dissipate the heat. If a positive air gap is used then there would be spaces between the deposited layers. Therefore, the semi-molten material extruded from the nozzle will spread between the spaces and it flows in and fills up the numerous voids making the part structurally stronger.

3.3 Effect of Raster Angle Raster angle shows low influence on storage modulus, loss modulus, and mechanical damping as shown in Fig. 6(a)-(c). The increase in raster angle from low level (0) to higher level (90) shows that there is little decrease in storage modulus and loss modulus but mechanical damping remains almost same. This could be related to the following facts. Firstly, a lower raster angle reduces the rasters length, which leads to a decrease

Fig. 6

Main effect plots showing effect of processing parameters on (a) storage modulus, (b) loss modulus, and (c) mechanical damping

in distortion, and hence a stronger interlayer bonding is obtained. Secondly, a higher raster angle creates the large voids between perimeter walls and deposited rasters, resulting in incomplete filling, hence poor dynamic mechanical performance of the built parts. Figure 8 shows the scanning electron microscopy (SEM) observation of the specimens manufactured with raster angle of 90. It is observed from Fig. 8 that there are sharp turns in the manufactured specimen at higher raster angle, which causes incomplete filling. Therefore, the part processed at lower raster angle is expected to have a dense structure and superior dynamic mechanical properties with less porosity. The influence of raster angle on the functionality of the manufactured part is illustrated further in Fig. 9. Here, six additional samples were manufactured. The first three samples were manufactured with raster angle of 0. The remaining three samples were manufactured with raster angle of 90. Other process parameters were kept at their original levels. The average value for each mechanical property was taken from a set of three values obtained from tested specimens. It can be

seen from Fig. 9 that if the sample is processed with raster angle of 0 and is subjected to the tension load, then the load requires separating and fracturing several number of layers, rasters, fibers, and polymer chain molecules. Hence, the samples processed with raster angle of 0 exhibit higher mechanical properties and resistance to deformation. However, if the sample is processed with raster angle of 90, the sample quickly reaches the ultimate failure and breaks. This is due to the fact that the separation and fracture occur at the interface of rasters (see Fig. 9) from that to which it is bonded.

3.4 Effect of Build Orientation It can be seen from Fig. 6(a)-(c) that influence of build orientation on the storage modulus, loss modulus, and mechanical damping is similar to what is observed for the effect of raster angle on these properties. However, its effect on the mechanical performances is more visible. As the build orientation changes from 0 to 90, the dynamic mechanical

Journal of Materials Engineering and Performance

3.5 Effect of Road Width Results from this study indicate that road width has marginal effect on all dynamic mechanical properties as shown in Fig. 6(a)-(c), but its effect on mechanical damping is noticeably visible than on storage modulus and loss modulus. From Fig. 6(a)-(c), it can be seen that both storage modulus and loss modulus are slightly higher at the lowest value of road width (0.4572 mm). This is due to the fact that the lowest value of road width provides a finer path to fill interior rasters, resulting in reducing the chances of incomplete filling. Conversely, mechanical damping increases with the increase in road width from 0.4572 to 0.5782 mm. This is because the higher value of road width (0.5782 mm) provides uniform thermal gradients and thermal cycles, resulting in minimum distortion. Moreover, thicker road widths create thick rasters, which improves the strength of the manufactured part to resist deformation, and hence more energy can be absorbed.

3.6 Effect of Number of Contours

Fig. 7 SEM micrographs showing (a), pinholes in the specimen fabricated with layer thickness of 0.178 mm, and (b) tear and delamination on the specimen processed with layer thickness of 0.127 mm

Figure 6(a)-(c) shows that the number of contours has a strong effect on all dynamic mechanical responses. It is obvious from the effect plots shown in Fig. 6(a)-(c) that increasing the number of contours from its lowest level to highest level would result in a significant increase in the storage modulus, loss modulus, and mechanical damping. This is due to the fact that using higher value of the number of contours results in decreasing the number of rasters, raster length, and air voids (see Fig. 10). Therefore, the load applied on the part is carried out by the contours, rather than by rasters, which leads to greater mechanical performance of manufactured part. This can be seen in Fig. 11, where SEM micrograph of the specimen manufactured with higher number of contours is noticed. This figure indicates that dense filling in the specimen and uniform structure can be obtained using higher number of contours. Furthermore, using a higher number of contours helps to distribute the stress applied on wider area between fibers throughout the part that helps in preventing further stress to cause any damage. This can significantly reduce the chance of micro-crack formation in the processed part and helps to prevent premature failure of manufactured parts.

3.7 Interaction Effects

Fig. 8 SEM micrograph showing sharp turns and incomplete filling in the specimen processed with raster angle of 90

properties become weaker. This is due to the fact that the number of layers required for manufacturing the part increases with vertically oriented part. Thus, distorted growth becomes a dominant factor resulting in poor performance characteristics.

Journal of Materials Engineering and Performance

Significant interaction effects were observed between different process parameters on the dynamic mechanical properties. Figure 12(a)-(c) shows the contour plot of the effects of air gap and number of contours on the dynamic mechanical behavior (storage modulus, loss modulus, and mechanical damping). These results demonstrate that higher dynamic mechanical properties are obtained with higher number of contours and lower air gap. The findings are in agreement with the above findings presented in section 3.6. These interaction effects provide additional validation of the results obtained from this study. Figure 13(a)-(c) shows the contour plot of the effects of raster angle and layer thickness on the storage modulus, loss modulus, and mechanical damping. It can be seen that the dynamic mechanical properties of the processed part are increased by increasing the layer thickness and decreasing the raster angle. The main reason is thick layers and rasters result in minimum number of layers and shorter rasters, which leads to minimum interlayer distortion. Therefore, according to this

Fig. 9

Fig. 10

Failure on the part manufactured with raster angle of 0 and raster angle of 90

Effect of number of contours on dynamic mechanical performance

contour plot, an optimum (maximum) dynamic mechanical performance setting is obtained with the layer thickness of 0.3302 mm and raster angle of 0.

4. Optimization In this study, the desirability function was used to determine the optimal process conditions to maximize the dynamic mechanical properties. Desirability function approach was used to identify the optimal settings of input variables settings that simultaneously optimize a set of responses. In this regard, the developed regression models given in Eq 7-9 were used to obtain the optimum settings for the FDM parameters and along with corresponding lower value, target value, upper value, weight, and level of importance of the responses as shown in Table 5. The weight reflects the shape of desirability function,

Fig. 11 SEM micrograph showing dense structure for the specimen processed with 10 contours

Journal of Materials Engineering and Performance

Fig. 12 Contour plots showing the interaction effects between air gap and number of contours on: (a) storage modulus, (b) Loss modulus, and (c) mechanical damping

Fig. 13 Contour plots showing the interaction effects between layer thickness and raster angle on: (a) storage modulus, (b) Loss modulus, and (c) mechanical damping

Journal of Materials Engineering and Performance

Table 5 Constraints of input parameters and responses Name A:Layer thickness B:Air gap C:Raster angle D:Build orientation E:Road width F:Number of contours Storage modulus Loss modulus Mechanical damping

Goal

Lower limit

Upper limit

Lower weight

Upper weight

Restricted to 0.127, 0.1778, 0.254, 0.3302 Is in range Is in range Is in range Is in range Is in range Maximize Maximize Maximize

0.127 0 0 0 0.4572 1 444.33 43.836 0.72875

0.3302 0.5 90 90 0.5782 10 1472.35 170.493 1.13223

1 1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1 1

which fine-tunes how the optimization process searches for the optimal settings. The ‘‘importance’’ of a goal is used to specify comparative importance of responses (Ref 22). The algorithm calculates the optimal settings for the input parameters along with desirability values to show how well those settings meet the responses requirements. The optimization is accomplished by obtaining the individual desirability (d) to maximize of the each response using Eq 10, and then combined the individual desirability to obtain the overall desirability (D) using Eq 11.

d¼

Fig. 14 Perturbation plot showing the sensitivity of the response using the desirability function

Fig. 15

8 < 0 :

 yL r T L

1

yT

ðEq 10Þ

where L is the lower limit, T is the target, y is response, and r is the desirability weight.

Ramp function plot for optimized parameters

Journal of Materials Engineering and Performance

•

•

•

Fig. 16 Contour plot showing composite desirability for the best solution

1 n

D ¼ ðd1  d2  d3 . . .  dn Þ ¼

n Y

•

!1n di

;

ðEq 11Þ

i¼1

where D is the overall desirability, n is the number of responses in the measure, and d1 ; d2 ; d3 . . . dn are individual desirability for a single response. Perturbation plot presented in Fig. 14 shows the response sensitivity through the desirability function. A higher curvature by a factor shows that the response is sensitive to modification of the factor. This figure demonstrates that the responses are sensitive to layer thickness, air gap, raster angle, and number of contours, but responses are less sensitive to the modification of build orientation and road width under optimal conditions. The optimization plot presented in Fig. 15 was generated using DX software version 9. This figure reveals that the optimal process parameters to maximize all dynamic mechanical responses are layer thickness of 0.3302 mm, air gap of 0.00 mm, raster angle of 0.0, build orientation of 0.0, road width of 04572 mm, and 10 contours. After optimizing of each level of the process parameters and its corresponding response value, a confirmation experiment was executed using the optimized process settings. The storage modulus of 1468.33 MPa, loss modulus of 166.98 MPa, and mechanical damping of 1.10 were obtained from the confirmation experiment. These findings are in good agreement with the estimated values shown in Fig. 15 obtained from desirability function. The composite desirability for this optimal setting is 0.977 (see Fig. 16), which means that the optimal process settings are ideal.

5. Concluding Remarks In this study, a functional relationship between process parameters and dynamic mechanical properties for the FDM process has been established through fractional factorial design and the following conclusions were obtained:

Journal of Materials Engineering and Performance

•

The FDM process parameters air gap and number of contours show a significant effect on the storage modulus and loss modulus. A gradual increase in storage modulus has been observed with the decrease in air gap and increase in the number of contours. However, an increase in layer thickness and decrease in raster angle, build orientation, and road width help to improve storage modulus and loss modulus. In the case of mechanical damping, the same trend, which is similar to storage modulus and loss modulus, was observed in terms of the effect of air gap and number of contours. However, the layer thickness, build orientation, and road width show a significant effect, where with the increase in layer thickness and road width and decrease in build orientation, mechanical damping increases. But raster angle has no effect on the mechanical damping. The SEM morphology shows that inappropriate selection of process parameters can cause incomplete filling, distortion, voids, and delamination, which eventually leads to poor mechanical performance of the manufactured part. The SEM morphology also shows that the lowest value of layer thickness can cause pinholes and voids in the part. This is due to the fact that the large number of layers is needed in case of using lower values of layer thickness, which requires an increase in thermal cycles, non-uniform temperature, and cooling time. The optimization results revealed that the optimal process conditions to maximize the storage modulus, loss modulus, and mechanical damping simultaneously are layer thickness of 0.3302 mm, zero air gap, raster angle of 0.0, build orientation of 0.0, road width of 0.4572 mm, and 10 contours.

References 1. O.A. Mohamed, S.H. Masood, and J.L. Bhowmik, Optimization of Fused Deposition Modeling Process Parameters: A Review of Current Research and Future Prospects, Adv. Manuf., 2015, 3, p 42–53 2. A. Gebhardt, Understanding Additive Manufacturing: Rapid Prototyping-Rapid Tooling-Rapid Manufacturing, Carl Hanser Verlag GmbH Co KG, Munchen, 2012 3. K.J. Christiyan, U. Chandrasekhar, K. Venkateswarlu, A Study on the Influence of Process Parameters on the Mechanical Properties of 3D Printed ABS Composite, IOP Conf. Series: Materials Science and Engineering, Vol 114, IOP Publishing, 2016, p 012109 4. A.K. Sood, R.K. Ohdar, and S.S. Mahapatra, Experimental Investigation and Empirical Modelling of FDM Process for Compressive Strength Improvement, J. Adv. Res., 2012, 3, p 81–90 5. G.C. Onwubolu and F. Rayegani, Characterization and Optimization of Mechanical Properties of ABS Parts Manufactured by the Fused Deposition Modelling Process, Int. J. Manuf. Eng., 2014, 2014, p 1–13 6. G. Percoco, F. Lavecchia, and L.M. Galantucci, Compressive Properties of FDM Rapid Prototypes Treated with a Low Cost Chemical Finishing, Res. J. Appl. Sci. Eng. Technol, 2012, 4, p 3838–3842 7. S.H. Masood, K. Mau, W. Song, Tensile Properties of Processed FDM Polycarbonate Material, Materials Science Forum, Trans Tech Publ, 2010, 654-656, p 2556-2559 8. A. Lanzotti, M. Grasso, G. Staiano, M. Martorelli, E. Pei, and R.I. Campbell, The Impact of Process Parameters on Mechanical Properties of Parts Fabricated in PLA with an Open-Source 3-D Printer, Rapid Prototyp. J., 2015, 21, p 604–617 9. I. Durgun and R. Ertan, Experimental Investigation of FDM Process for Improvement of Mechanical Properties and Production Cost, Rapid Prototyp. J., 2014, 20, p 228–235

10. F. Go´rski, W. Kuczko, R. Wichniarek, A. Hamrol, Computation of Mechanical Properties of Parts Manufactured by Fused Deposition Modeling Using Finite Element Method, 10th International Conference on Soft Computing Models in Industrial and Environmental Applications, Vol 368, Springer, 2015, p 403-413 11. Y. Zhang and K. Chou, A Parametric Study of Part Distortions in Fused Deposition Modelling Using Three-Dimensional Finite Element Analysis, Proc. Inst. Mech. Eng. Part B J. Eng. Manuf., 2008, 222, p 959– 968 12. D. Impens and R. Urbanic, Assessing the Impact of Post-Processing Variables on Tensile and Compression Characteristics for 3D Printed Components, IFAC-PapersOnLine, 2015, 48, p 652–657 13. A. Arivazhagan, S. Masood, I. Sbarski, Dynamic mechanical analysis of FDM rapid prototyping processed polycarbonate material, Proceedings of the 69th Annual Technical Conference of the Society of Plastics Engineers (ANTEC 2011), Boston, Massachusetts, 2011, p 950-955 14. A. Arivazhagan and S. Masood, Dynamic Mechanical Properties of ABS Material Processed by Fused Deposition Modelling, Int. J. Eng. Res. Appl., 2012, 2, p 2009–2014

15. A.K. Sood, A. Equbal, V. Toppo, R. Ohdar, and S. Mahapatra, An Investigation on Sliding Wear of FDM Built Parts, CIRP J. Manuf. Sci. Technol., 2012, 5, p 48–54 16. S.K. Panda, S. Padhee, S. Anoop, S. Anoop Kumar, and S. Mahapatra, Optimization of Fused Deposition Modelling (FDM) Process Parameters Using Bacterial Foraging Technique, Intell. Inf. Manag., 2009, 1, p 89–97 17. K.P. Menard, Dynamic Mechanical Analysis: a Practical Introduction, CRC Press, Boca Raton, 2008 18. T.G. Mezger, The Rheology Handbook: For Users of Rotational and Oscillatory Rheometers, Vincentz Network GmbH & Co KG, Hannover, 2006 19. A. Seidel, Characterization and Analysis of Polymers, Wiley, Hoboken, 2008 20. D.C. Montgomery, Design and Analysis of Experiments, Wiley, Hoboken, 2008 21. R. Pannerselvam, Design and Analysis of Experiments, PHI, Learning Pvt. Ltd., New Delhi, 2012 22. L. Eriksson, Design of Experiments: Principles and Applications, MKS Umetrics AB, Malmo, 2008

Journal of Materials Engineering and Performance