**2.2 Specimen preparation and equipment**

This project included five different mechanical tests: tension, compression, flexural (3-point bend), impact, and tension-tension fatigue. Three unique specimen designs were required. The tension, flexural, and fatigue specimens were all thin rectangular slabs (Figure 2a) fabricated to be 190.5 mm long, 12.7 mm wide, and 2.6 mm thick in accordance with ASTM D3039 (ASTM, 1998), ASTM D790 (ASTM, 2007), and ASTM D3479 (ASTM, 2007a) standards respectively. Compression specimens were cylindrical and fabricated with dimensions conforming to the ASTM D695 standard (ASTM, 1996). Each cylinder was 25.4 mm long and 12.7 mm diameter (Figure 2b). Impact specimens were fabricated with dimensions conforming to the ASTM D256 standard (ASTM, 2010). The geometry was a v-notched rectangular block of 63.5 mm long, 25.4 mm wide, and 25.4 mm thick (Figure 2c). The vnotch was modeled within the computer solid model of the specimen and was produced directly on the FDM machine.

Fig. 2. Specimen geometries associated with each test

All FDM specimens were fabricated with a Stratasys Vantage-*i* machine. Solid models were first created using Pro/Engineer® software, and then tessellated and exported in STL format. Digital models were then sliced using the Vantage machine's Insight software, and layer extrusion tool paths were generated, i.e. raster patterns used to fill interior regions of each layer, to represent the four different fiber orientations studied in each test.

Layered specimens were all fabricated in a build orientation that aligned the minimum part dimension with the z-axis of the machine, i.e. perpendicular to the build platform. In this orientation, five to ten replicate specimens were built with each of four different raster patterns relative to the part loading direction, for each of the five different tests completed. The four raster orientations included: (a) longitudinal or 0º, i.e. rasters aligned with long dimension of the specimen, (b) diagonal or 45º, i.e. rasters at 45º to the long dimension of the specimen, (c) transverse or 90º, i.e. rasters perpendicular to long dimension of the specimen, and (d) default or +45º/-45º criss-cross, i.e. representing the machine's default raster orientation (Figure 3).

Fig. 3. Four different raster orientations investigated

All FDM specimens were built while holding all other machine process settings at the recommended or default values displayed in Table 1.


Table 1. Fixed FDM process settings

162 Mechanical Engineering

family. It is a common thermoplastic that is formed by dissolving butadiene-styrene copolymer in a mixture of acrylonitrile and styrene monomers, and then polymerizing the monomers with free radial initiators (Odian, 2004). The result is a long chain of polybutadiene crisscrossed with shorter chains of poly(styrene-co-acrylonitrile). The advantage of ABS is that it combines the strength and rigidity of the acrylonitrile and styrene polymers with the toughness of the polybutadiene rubber. The proportions can vary from 15 to 35% acrylonitrile, 5 to 30% butadiene, and 40 to 60% styrene. In this study, the resulting composition was 90-100% acrylonitrile/butadiene/styrene resin, with 0-2%

This project included five different mechanical tests: tension, compression, flexural (3-point bend), impact, and tension-tension fatigue. Three unique specimen designs were required. The tension, flexural, and fatigue specimens were all thin rectangular slabs (Figure 2a) fabricated to be 190.5 mm long, 12.7 mm wide, and 2.6 mm thick in accordance with ASTM D3039 (ASTM, 1998), ASTM D790 (ASTM, 2007), and ASTM D3479 (ASTM, 2007a) standards respectively. Compression specimens were cylindrical and fabricated with dimensions conforming to the ASTM D695 standard (ASTM, 1996). Each cylinder was 25.4 mm long and 12.7 mm diameter (Figure 2b). Impact specimens were fabricated with dimensions conforming to the ASTM D256 standard (ASTM, 2010). The geometry was a v-notched rectangular block of 63.5 mm long, 25.4 mm wide, and 25.4 mm thick (Figure 2c). The vnotch was modeled within the computer solid model of the specimen and was produced

All FDM specimens were fabricated with a Stratasys Vantage-*i* machine. Solid models were first created using Pro/Engineer® software, and then tessellated and exported in STL format. Digital models were then sliced using the Vantage machine's Insight software, and layer extrusion tool paths were generated, i.e. raster patterns used to fill interior regions of

(a) (b) (c)

Layered specimens were all fabricated in a build orientation that aligned the minimum part dimension with the z-axis of the machine, i.e. perpendicular to the build platform. In this orientation, five to ten replicate specimens were built with each of four different raster patterns relative to the part loading direction, for each of the five different tests completed. The four raster orientations included: (a) longitudinal or 0º, i.e. rasters aligned with long dimension of the specimen, (b) diagonal or 45º, i.e. rasters at 45º to the long dimension of the specimen, (c) transverse or 90º, i.e. rasters perpendicular to long dimension of the specimen, and (d) default or +45º/-45º criss-cross, i.e. representing the machine's default raster

each layer, to represent the four different fiber orientations studied in each test.

mineral oil, 0-2% tallow, and 0-2% wax.

directly on the FDM machine.

orientation (Figure 3).

Fig. 2. Specimen geometries associated with each test

**2.2 Specimen preparation and equipment** 

In order to measure the reference strength and behaviour of the ABS filament material, for comparisons with the layered parts, additional specimens were fabricated by injection molding for the same five tests. Aluminium molds for each of the three previously described geometries (Figure 2) were designed using Pro/Engineer® software, and manufactured on a Haas VF-1 CNC machining center. Mold cavity dimensions were the same as those described for the FDM specimens, with slight increases to compensate for the shrinkage of molded ABS at approximately 0.005 cm/cm. Parting lines and runners were located with an effort to avoid potential stress concentrations or anomalies in the resulting specimens that might affect test results. All molded specimens were fabricated from the same material as the layered models by feeding the FDM-ABS filament into a polymer granulator and cutting it into pellets of 3-5 mm in length. The pellets were then fed into the hopper of a Morgan Press G-100T injection molding machine. Molding parameters were set to the recommended values for ABS plastic, including nozzle temperature of 270 °C, mold preheat temperature of 120° C, clamping force of 71 kN (16,000 lb), and injection pressure of 41 MPa (6000 psi). Ten replicate specimens were molded for each of the five tests.

Tensile, compressive, flexural, and tension-fatigue tests were performed on an Instron model 3366 dual column uniaxial material testing with .057 micron displacement precision and up to 0.001 N force accuracy. The machine has a 10kN load force capacity. Impact strength was studied on a TMI impact tester. Resulting fracture surfaces were subsequently prepared by gold sputtering and analyzed with a JSM 500-type JEOL Scanning Electron Microscope (SEM).

Anisotropic Mechanical Properties of ABS Parts Fabricated by Fused Deposition Modelling 165

Source DF SS MS F P Raster Angle 3 363.991 121.330 460.98 0.0001

Table 3. ANOVA results comparing mean ultimate tensile strengths of 4 raster orientations

Further analysis in the form of post hoc comparisons was performed to determine which raster orientations differed in mean UTS. Tukey's method (Montgomery, 2009), creating a set of 95% simultaneous confidence intervals for the difference between each pair of means, indicated that the difference was significant for all pairwise comparisons of mean UTS values (Table 4). The difference between the mean UTS of the longitudinal rasters (25.72) and that of the transverse rasters (14.56) was the most significant. These results confirm that raster orientation has a significant effect on the tensile strengths of the FDM specimens. Tensile strength is thus verified to be affected by the directional processing and subsequent

> **Difference of Mean UTS (i-j)**

Transverse 11.16\* 10.235 12.093 Default 6.36\* 5.438 7.296

Default -3.14\* -4.07 -2.212

Longitudinal 45-Degree 9.50\* 8.579 10.437

45-Degree Transverse 1.66\* 0.727 2.585

Transverse Default -4.8\* -5.726 -3.868

The quantitative data analysis was followed by detailed physical inspection of the specimens at both macro and microscopic levels. Macroscopically, the fracture patterns of the specimens varied somewhat as a function of the raster orientation of the twodimensional layers and the resulting weakest path for crack propagation (Figure 5). The 90º specimens failed in the transverse direction and the 45º specimens failed along the 45º line. The 0º specimens failed primarily in the transverse direction, although there was some fiber pullout and delamination intermittently evident as well. The +45º/-45º specimens broke at intersecting fracture paths along ±45º, resulting in a saw-tooth fracture pattern across the specimen width. It is likely that fracture paths controlled by weak interlayer bonding are affected by the residual stresses that result from the volumetric shrinkage of the polymer layers during solidification and cooling. In addition, interlayer porosity and air gaps serve to reduce the actual load-bearing area across the layers, providing an easy fracture path.

In specimens with the longitudinal (0°) raster orientation, the molecules tend to align along the stress axis direction. This produces the strongest individual two-dimensional layers subjected to tension loading. During the testing of these specimens, stress whitening due to

Table 4. Post hoc Tukey HSD multiple comparisons of mean tensile strengths

**95% Confidence Interval** 

**Upper Bound** 

**Lower Bound** 

Error 16 4.211 0.263 Total 19 368.202

directionality of the polymer molecules, signifying an anisotropic property.

**Raster Orientation (j)** 

\* Mean difference is significant at the 0.05 level

**Raster Orientation (i)** 
