APEM jowatal Advances in Production Engineering & Management Volume 13 | Number 3 | September 2016 | pp 227-238 http://dx.doi.Org/10.14743/apem2016.3.223 ISSN 1854-6250 Journal home: apem-journal.org Original scientific paper Investigation of dynamic elastic deformation of parts processed by fused deposition modeling additive manufacturing Mohamed, Omar A.a*, Masood, Syed H.a, Bhowmik, Jahar L.b aDepartment of Mechanical and Product Design Engineering, Swinburne University of Technology, Hawthorn, Australia bDepartment of Statistics, Data Science and Epidemiology, Swinburne University of Technology, Hawthorn, Australia A B S T R A C T A R T I C L E I N F O Fused deposition modeling (FDM) has been recognized as an effective technology to manufacture 3D dimensional parts directly from a digital computer aided design (CAD) model in a layer-by-layer style. Although it has become a significantly important manufacturing process, but it is still not well accepted additive manufacturing technology for load-carrying parts under dynamic and cyclic conditions due to many processing parameters affecting the part properties. The purpose of this study is to characterize the FDM manufactured parts by detecting how the individual and interactive FDM process parameters will influence the performance of manufactured products under dynamic and cyclic conditions. Experiments were conducted through fractional factorial design and artificial neural network (ANN). Effect of each parameter on the dynamic modulus of elasticity was investigated using analysis of variance (ANOVA) technique. Furthermore, optimal processing parameters were determined and validated by conducting verification experiment. The results showed that both ANN and fractional factorial models provided good quality predictions, yet the ANN showed the superiority of a properly trained ANN in capturing the nonlinear relationship of the system over fractional factorial for both data fitting and estimation capabilities. © 2016 PEI, University of Maribor. All rights reserved. Keywords: Additive manufacturing Fused deposition modeling (FDM) Dynamic modulus of elasticity Fraction factorial design Artificial neural network (ANN) Process parameters Analysis of variance (ANOVA) *Corresponding author: Omar.Ahmed.Mohamed@outlook.com (Mohamed, Omar A.) Article history: Received 18 June 2016 First Revision 22 August 2016 Second Revision 28 August 2016 Accepted 29 August 2016 1. Introduction Fused deposition modeling (FDM) process is the most popular Stratasys-patented additive manufacturing technology. FDM is gaining importance in many manufacturing applications due to its ability to create complex prototypes without requiring any tools [1]. This process builds 3D shapes from a digital CAD file in a layer-by-layer format from the bottom by melting and extruding a fine filament of thermoplastic from the extrusion nozzle onto a base. The nozzle moves horizontally and vertically over the build table to translate the dimensions of part into the X, Y and Z axes. Over the past decade, FDM process has gained increasing attention in the field of 3D manufacturing products. Although FDM has become a more sophisticated and the range of available materials continuing to grow, the application of this process in various industries is still not yet a fully accepted as a mature technique due to lower mechanical performance and performance of the fabricated parts compared with traditional manufacturing processes such as sheet metal 227 Mohamed, Masood, Bhowmik forming, thermoforming and injection molding. The main reason for poor mechanical performance of FDM built parts is the existence of a great number of intervening processing conditions affecting the overall part quality[1]. For instance, incorrect settings of operating parameters can cause defects on the manufactured products, such as void structures. Hence, it is essential to understand and optimize the impact of operating conditions during on the processed prototypes. During last couple of decade's extensive research has been carried out with limited success on optimizing FDM operating parameters for various quality characteristics such as mechanical properties, surface roughness, build quality and dimensional accuracy. For example, Wang et al. [2] reported that the mechanical performance can highly affected by part direction. This study also revealed that the highest mechanical strength can be obtained when the part was manufactured with minimum z-height. Rayegani and Onwubolu [3] have carried out an experiment on the impact of FDM operating conditions on mechanical strength of build parts. The results from this study have shown that small road width, negative air gap and zero build direction can improve the mechanical strength significantly. Sood et al. [4] concluded that thick layers and rasters with zero raster to raster air gap improve the mechanical characteristics significantly. Chris-tiyan et al [5]reported that using low printing speed and low layer thickness can effectively improve the mechanical performance of FDM built prototypes. Impens and Urbanic [6] investigated the influence of post-processing settings on the mechanical characteristics for built parts. They found that build direction is the key factor in optimizing tensile and compression strengths for processed parts. Recently, Lanzotti et al. [7] studied the impact of process parameters like infill direction, slice height and perimeters on the prototype strength. This study reported that high variation in the mechanical strength can be noticed by changing in the level of each processing parameters. Very few studies have been made on the investigation of the effect of processing parameters on mechanical properties under cyclic loading conditions. For example, Ari-vazhagan et al [8] examined the influence of built style, road width, and raster pattern on the dynamic mechanical performance of polycarbonate manufactured part. Arivazhagan et al [9]conducted similar study on the effect of FDM operating conditions but on the part made by ABS material. In both studies, they conducted their experiments based on the trial and error approach. Their results indicated that the maximum dynamic performance can be obtained by using solid build style, 45° raster pattern and 0.454 mm road width. Although during last decade a remarkable progress has been made in FDM process parameters optimization technique, but most of the existing literature focused only on improving the mechanical properties under static leading conditions. In fact, the parts manufactured by the FDM process are also subject to vibratory and cyclic conditions for long-term prediction with wide range of temperatures. There are two studies done so far on dynamic mechanical properties. However, they are expensive due to the use of one-factor-at-a-time (OFAT) method as well as they are limited in terms of the number of processing parameters being investigated and type of dynamic mechanical property observed. OFAT method cannot lead to optimal process settings and the relationships between the processing conditions and dynamic mechanical response using this approach are still unclear. This paper differs from all previous studies in several ways. Firstly, unlike previous studies, which focused on the effect of FDM processing parameters on the static mechanical properties of the manufactured parts, this study examines the effect of FDM process conditions on the dynamic mechanical properties that resulted in understanding the material behaviour under cyclic loading conditions. Secondly, unlike most previous studies that aimed at investigating the influence of only few FDM process parameters, this paper considers the effect of six FDM processing parameters including a new variable - number of contours - which was not studied in the published literature before. Finally, unlike most previous studies, this study explores whether there is a significant relationship between the FDM process parameters and dynamic mechanical property, namely dynamic modulus of elasticity using fraction factorial design, regression analysis and artificial neural network (ANN). Results show that optimal process parameters lead to achieve desired dynamic modulus of elasticity of FDM produced part. Results obtained from this study would be useful for industry application and would help to produce the end user products with desired dynamic mechanical properties. It also can be used as a guide for planning and carrying out future studies. 228 Advances in Production Engineering & Management 11(3) 2016 Investigation of dynamic elastic deformation of parts processed by fused deposition modeling additive manufacturing 2. Materials and methods The experiments in this study were designed and performed using fraction factorial design. Fraction factorial design experimental design is commonly used to determine the most critical factors in the early stages of experimental work, when several process parameters are likely to be investigated as well as when the knowledge about the process is usually unavailable[10, 11]. This study used the stipulated conditions according to the fraction factorial design to plan the experiments. A total of 16 experiments were conducted at two levels of each input parameter. Two level fraction factorial experiment involves an experimental design in which each parameter is investigated at two levels. The early stages of experimental work and investigation usually involve the study of a large number of parameters to determine the vital parameters important for the system. Two level fraction factorial design is used in these stages to find out unnecessary factors so that attention can then be made only on the critical factors. The data were analysed using STATISTICA software. The experimental design used in this study considered the following processing parameters to investigate their effect on the dynamic modulus of elasticity; layer thickness (A), air gap (B), raster angle (C), build orientation (D), road width (E) and the number of contours (F). The selected process parameters and their levels are presented in Table 1 and they are selected according to the previous studies and FDM machine manufacturer (Stratasys) guide. The FDM build parameters are presented in Fig. 1. Table 1 FDM process parameters and their levels Factors Units Code Levels Low High Layer thickness Air gap Raster angle Build orientation Road width Number of contours mm mm deg deg mm A B C D E F 0.127 0 0 0 0.4572 1 0.3302 0.5 90 90 0.5782 10 A total of 16 samples having dimension of 35 (length) mm x 12.5 mm (width) x 3.5 mm (thickness) were fabricated by FDM Fortus 400 as per designed plan presented in Table 2 and tested according to ASTM D5418 [12] and TA instrument manufacturer recommendations [13]. All samples are made by PC-ABS material which has amorphous structures. Dynamic modulus of elasticity is a viscoelastic property that exhibit both viscous and elastic behaviors which is present in the material or manufactured part during undergoing deformation. It is the ratio of peak dynamic stress to peak dynamic strain under vibration and harmonic loading. Therefore, dynamic modulus of elasticity measures the sample and material resistance to deformation [14]. Layer Thickness *Sf L V Z Y, X "I" i 90° Road width Raster angle Number of contours Air gap Build Orientations Tool Path Parameters Fig. 1 FDM build parameters Advances in Production Engineering & Management 11(3) 2016 229 Mohamed, Masood, Bhowmik Fig. 2 Schematic illustration of dynamic mechanical test The dynamic mechanical response in terms of dynamic modulus of elasticity of the 16 samples was measured using 2980 Dynamic Mechanical Instrument in the bending mode with single cantilever. Dynamic mechanical measurement was done with single frequency of 1 Hz with a heating rate of 3 °C /min, oscillation amplitude of 15 [im, and the temperature ranges between 35-170 °C with soaking time of 5 min. The stress-strain curve, which was generated by dynamic mechanical machine and analysed by Thermal Advantage Software, has been used to determine the maximum dynamic modulus of elasticity for each experimental run according to fraction factorial design matrix plan. The average of the maximum values of dynamic modulus of elasticity was taken from a set of tested samples. The experimental design plan in terms of coded parameter with the measured dynamic modulus of elasticity is presented in Table 2. Table 2 Experimental design matrix Run A B C D E F Dynamic modulus of elasticity (MPa) 1 0.127 0.5 90 0 0.4572 1 4.255 2 0.127 0.5 90 90 0.4572 10 11.028 3 0.127 0 90 0 0.5782 10 12.339 4 0.127 0 0 90 0.4572 10 12.946 5 0.127 0.5 0 90 0.5782 1 5.542 6 0.127 0.5 0 0 0.5782 10 13.056 7 0.127 0 90 90 0.5782 1 10.881 8 0.127 0 0 0 0.4572 1 12.228 9 0.3302 0.5 90 0 0.5782 1 4.732 10 0.3302 0 90 0 0.4572 10 14.287 11 0.3302 0 90 90 0.4572 1 12.829 12 0.3302 0.5 0 90 0.4572 1 4.240 13 0.3302 0 0 90 0.5782 10 12.771 14 0.3302 0.5 90 90 0.5782 10 11.504 15 0.3302 0 0 0 0.5782 1 12.054 16 0.3302 0.5 0 0 0.4572 10 11.753 3. Results and discussion The relationships between measured dynamic modulus of elasticity and the FDM process parameters were developed by fitting the data in a two-factor interaction (2FI) model presented in Eq. 1, where, Y is the predicted response (dynamic modulus of elasticity), is a constant intercept, Pi is the coefficient for the linear terms, is the interaction coefficient, Xt and Xj are the coded factors, and and £ is the random error term. 230 Advances in Production Engineering & Management 11(3) 2016 Investigation of dynamic elastic deformation of parts processed by fused deposition modeling additive manufacturing 6 6 Y = p0 + YJ Mi+ ZX XiXj + £ [1) ¿=i i F Model 182.11 6 30.35 91.25 < 0.0001 A 0.22 1 0.22 0.67 0.4327 B 73.21 1 73.21 220.09 < 0.0001 C 0.47 1 0.47 1.40 0.2664 F 67.75 1 67.75 203.67 < 0.0001 AC 3.80 1 3.80 11.44 0.0081 BF 36.66 1 36.66 110.22 < 0.0001 Residual 2.99 9 0.33 - - Total 185.11 15 - - - : = 98.38 % 'o, Adjusted R2 = 97.30 <6, Predicted R2 = 94.89 %, Adequate precision = 25.454 Advances in Production Engineering & Management 11(3) 2016 231 Mohamed, Masood, Bhowmik (a) [Standardized Effect| (b) Externally Studentized Residuals Fig. 3 (a) half normal probability plot of the standardized effects, and (b) normal probability plot, for dynamic modulus of elasticity Predicted vs. Actual Residuals vs. Run (a) Actual (b) Run Number Leverage vs. Run 1.000.600.600.400.20- Dynamic motfulus ol elasticity [MPa] ■ Í4.2S7 14.240 1 4 7 10 13 16 (c) Run Number Fig. 4 (a) predicted versus actual plot, (b) residual versus run number plot, and (c) leverage versus run number plot Effect of layer thickness (slice thickness) on the dynamic modulus of elasticity of the parts can be seen in Fig. 5. With the increase in slice thickness, dynamic modulus of elasticity of the part slightly increased. It is because as the layer thickness increases, it produces thick rasters with minimum number of layer. This leads to the improvement in dynamic mechanical properties of the built part. Nevertheless, if the part is fabricated with thin layers, there would be micro-voids and tear in a part surface (see Fig. 6). Thus the sample processed with thin layers exhibits lower mechanical performance. Fig. 5 reveals the influence of air gap on the dynamic modulus of elasticity. It is found that with an increase in air gap, dynamic modulus of elasticity decreases gradually. The main reason is that when the air gap increases, a close raster and deposited beads are generated, which leads to a dense structure resulting in improvement in dynamic modulus of elasticity of parts. 232 Advances in Production Engineering & Management 11(3) 2016 Investigation of dynamic elastic deformation of parts processed by fused deposition modeling additive manufacturing Fig. 5 Effect of various operating conditions on dynamic modulus of elasticity Fig. 6 Microstructure observation of the effect of thin layer on the properties of the manufactured part Fig. 5 also shows the impact of raster angle (raster pattern) on the dynamic modulus of elasticity on the samples built through FDM. It has been observed the dynamic modulus of elasticity for the manufactured part decreases with increasing raster angle from 0° to 90°. The main reason behind this phenomenon is that when the raster angle increases, the energy absorbed by the manufactured part decreases. This is due to the fact that at raster angle of 90° an adhesive failure occurs at the bonding interface level of the deposited layers (see Fig. 7). This leads to reduction in the dynamic modulus of elasticity of the processed part. Fig. 7 clearly shows the phenomena behind the influence of two raster's angles on the dynamic modulus of elasticity. Bending load Fixed sample Bending load Bending load Sample processed Sample processed with with raster an«;!« of 0 raster angle of 90° Fig. 7 Failure of different rasters under periodic bending load Advances in Production Engineering & Management 11(3) 2016 233 Mohamed, Masood, Bhowmik The effect of build orientation on dynamic modulus of elasticity is illustrated in Fig. 5. To acquire high deformation resistance for the fabricate part, it is preferable to manufacturing the part along the X-axis (0°) as this can greatly improve the curve definition for rasters, and can decrease stair-stepping effect. Fig. 5 indicates that the road width has no effect on the dynamic modulus of elasticity. Thus this factor has been removed from the regression model expressed which is by Eq. 2. However, in general it is advisable to use thin road width as thin road width provides finer raters and layers, which helps in filling more spaces on the part structure. Thus the built parts tend to have better mechanical properties, better dimensional accuracy and improved surface roughness. The effect of number of contours on dynamic modulus of elasticity is shown in Fig. 5. The results indicate that higher values of dynamic modulus of elasticity can be obtained by considering 10 contours. The maximum contour lines can guarantee elevated absorb and discharge energy levels and help the part to return to its original position after the stress is released. Because the reason for this improvement is maximum number of contours reduces the number of rasters, which helps to create the solid and dense structure (see Fig. 8) and hence increases the dynamic modulus of elasticity. Fig. 8 Microstructure observation of the effect of 10 contours on the properties of the manufactured part Fig. 9 portrays the dual influence of air gap and number of contours on dynamic modulus of elasticity at a constant level of the other processing parameters. It can be concluded that maximum dynamic modulus of elasticity is feasible with a combination of low air gap and higher number of contours. However, an interesting phenomenon can be noticed from Fig. 9 that using highest value of air gap along with maximum number of contours higher dynamic modulus of elasticity can still be obtained. This is because the part still has solid structure under this parametric combination, and hence this combination of process parameters helps to improve the mechanical properties while reducing the production cost as positive air gap minimizes the processing time. 234 Fig. 9 Combined effect of air gap and number of contours on dynamic modulus of elasticity Advances in Production Engineering & Management 11(3) 2016 Investigation of dynamic elastic deformation of parts processed by fused deposition modeling additive manufacturing 20.000 15.117 0.0000 1 0000 .127 .3302 0. -a—s—s—B- —a—B—a—B— -a—B—a—B- -Q ns "(ft D O 1 r* J/ jQ ra k. en