CAD tools are necessary to successfully utilize AM technologies. CAD tools were optimized to aid the requirements and limitations of traditional manufacturing methods. Most CAD tools are built with the assumption of homogenous part’s interior from both geometrical and materials point of views. An appropriate CAD tool for AM should allow the freedom to tailor the material’s distribution and composition inside the part. In addition, it should allow the part’s fabrication to be free of the constraints of traditional manufacturing methods. Thus, we optimize the topology of the part according to the loads and functionality instead of manufacturability. An example of used CAD tools was discussed in Srinivas Bhashyam et al. 2000 [14] and David W. Rosen 2007 [15] , where various approaches were suggested to build heterogeneous parts. Currently, there are some commercial CAD tools that support AM such as Creo 4.0^® Additive Manufacturing, which allows the fabrication of the parts and its interior parametrically and slice it for printing directly. Solid Thinking^® is an example of CAD software that allows topology optimization as per the loading and the user defined requirements.

Due to the nature of AM that builds the parts layer by layer, the mechanical properties of the part are a function of the material used and the processing parameters. Processing parameters include the building orientation, extrusion temperature, overlap percentage, infill patterns and many others. Therefore, the strength of the part and its stiffness depend on the strength of the fusion and bonding between the extruded infill and the air gaps separating them. This makes simulating FDM parts in FEA a challenging task. Until now, there are no available packages for simulating AM parts and research is still in progress. In literature, there are several approaches taken in simulating AM parts. The first approach is done by testing tensile specimens in all principle directions to collect the stiffness matrix of the material. Then using this stiffness matrix simulations can be performed and verified [16] . However, this method does not show stress concentrations or explain the failure patterns. In [17] , the FDM parts were simulated by adding air gaps according to their relative density to the raw filament. The gaps were made to represent the shape of the infill patterns. This explains the stress concentrations and the failure criteria. However, the part was still considered to be a continuum density instead of layers. In [18] , the simulation model was built by patterning cylinders between 400 - 2000 times to create the specimen’s gauge length. This method should represent the nature of FDM parts. However, it requires great computational power to simulate complex parts.

The material library for FDM is constantly increasing to improve the applicability of FDM [19] [20] . However, the available materials are usually limited to thermoplastics, composites with thermoplastic. That is because most commercial FDM extruders are limited to up to 300˚C while enabling rapid solidification so that the printed layers do not lose shape. There are some materials that have high glass transition temperatures such as PEEK. However, PEEK requires extrusion temperature above 300˚C, which most commercial FDM systems cannot produce except after modifying the extrusion head and replacing the thermoset with a thermocouple.

Consequently, when designing for FDM, designers are limited by the small range of materials. Currently, the most common FDM filaments are PLA, ABS and Nylon for printing parts while PVA and HIPS are the most common support filaments due to their dissolvability. New filaments that use TPE or TPU (Thermoplastic polyurethane) print parts with much higher elasticity than ABS and PLA. Composite filaments include carbon fiber, metal powder or wood fibers with PLA, Nylon [21] or ABS [22] . Among the most common three filaments, Nylon has the highest strength and flexibility, followed with PLA that has a medium strength and least flexibility [23] . Finally, ABS has the lowest strength accompanied with higher flexibility than PLA. ABS has the advantage of being dissolvable in acetone [24] , which gives the ability to chemically treat its surface for better finishing. Table 1 lists the conservative mechanical properties of the most common filaments. Because of the low melting and glass transition temperatures, most FDM commercial filaments are not suitable for high temperature applications above 200˚C.

The mechanical properties of the material degrade significantly near the glass transition temperatures as shown in an experiment conducted by the authors. The tensile test specimens of PLA material were created in both orientations; horizontally and vertically and tested at room temperature, 50˚C and 60˚C using a controlled oven chamber attached to the universal tensile testing machine. A total of 18 fabricated specimens were tested in addition to 9 filament (raw material) specimens. The FDM specimens were fused using 0.20 mm layer thickness, three shells, 0.50 mm raster width, an extrusion temperature of 200˚C, and 40 mm/s printing speed with a nozzle of 0.5 mm in diameter on Taz 6^® FDM system using Pro series PLA filament of 3.0 mm diameter. Figure 3 illustrates the dimensions of the fabricated samples. Although the specimens showed comparable properties at room temperature, at temperatures close to the glass transition, the mechanical properties dropped drastically as shown in Figure 4.

Figure 4 indicates the results for four cases, those are the samples fabricated horizontally relative to the loading direction, samples fabricated vertically relative to the loading direction, raw material filament, and the typical injection molding PLA. The modules of elasticity results show a significant reduction at 50˚C environmental temperature compared to room temperature. It is also noted that the injection molding PLA has a higher modulus of elasticity at room temperature than the rest three cases. In addition, at 60˚C the modulus of elastic for the AM PLA samples reduced to a negligible value compared to the room temperature and the 50˚C samples since 60˚C is very close to the glass transition temperature of PLA. We can see that FDM parts maintained their anisotropy even at elevated temperatures where parts fabricated horizontally with their layers parallel to the load show comparable stiffness and strength to the raw filament. On the other hand, vertically fabricated parts with their layers normal the load exhibit significantly lower stiffness. Finally, the injection modeling PLA exhibited slightly higher modules of elasticity than any FDM sample at room temperature.

One of the major current limitations of FDM, and AM in general, is the small build volume that the process permits. Though there are FDM printers with

larger building volumes, most commercial FDM printers have building volumes less than 30 cm × 30 cm × 30 cm. Similarly, most FDM printers have Cartesian coordinate axis. However, many FDM printers have Delta coordinate as shown in Figure 5. This main difference is in the distribution of the building volume. Typically, the Cartesian FDM printers have bigger building bed but smaller building dimension in the height direction, while Delta FDM printers have smaller bed area in comparison to their Z-axis or height. The distribution of the building volume is important as it can limit the possible orientations for a given part. In addition, one advantage of AM is to avoid unnecessary assemblies by printing components as one part allowing for components consolidation. However, the limited building volume might force portioning the design into smaller parts. Consequently, adding assembly steps to the manufacturing process. For materials prone to distortion and warpage, the utilization of the full bed might result with parts of poor quality due to the temperature variations along the edges of the heated bed and different cooling rate, especially in an open FDM system [26] .

Most FDM printers have its manufacturer nozzle positioning accuracy procedure. Typical values range between 10 - 12.5 microns for the XY plane and 1 - 2.5 microns in the Z direction [27] [28] . Some equipment has even smaller values [29] . This accuracy might not reflect exactly on the resolution of the resulted FDM parts since more factors affect the parts’ resolution; such as the nozzle diameter and the filament material properties combined with the processing parameters used. Typical nozzle diameters range between 0.30 - 1.00 mm with 0.40 and 0.50 mm nozzles being the most common.

Accordingly, designers should be aware of small details in their designs such as angles, fillets, curves, small holes, and thin walls, as they might not be fabricated as intended in the design. Childs and Juster [30] investigated the dimensional accuracy of many AM technologies and tested how well small features such as short linear shapes, thin walls and small holes can be produced using FDM. It was found that FDM was not able to fuse holes with diameter of 1 mm or thin walls of 0.5 mm. In Johnson et al. [31] , a benchmarking part was printed with small features such as holes, cylinders, and notches with dimensions no smaller than 1 mm. It was found that for small cylinders with 6 mm diameter the deviation was 0.26 mm while for the notch with 2 mm the deviation was 0.35 mm.

In conclusion, features with smaller dimension than 1 mm should be avoided as they might fail to be rendered. Features between 1 mm and 5 mm will have deviations between 0.15 mm - 0.35 mm. In addition, it is favorable to have thin walls with thicknesses equal to an integer multiplication of the nozzle diameter or raster width [32] . Larger value of raster width will build a part with a stronger interior. Smaller value will require less production time and material. The value of raster width is mainly a function of the nozzle diameter.

Overhangs features, shown in Figure 6(a), are a main aspect in FDM that designers should carefully consider during the conceptual phase of CAD model. In general, there is a limit to the length of unsupported depositions that can be printed using FDM [5] . In the work of Belter and Dollar [33] , 30˚ over-hanged features with length of 3 mm were successful printed with no support. In Johnson et al. [31] work, longer overhangs were successfully printed without support as well. However, the overhangs features were fabricated with deviated angles. For example, the overhang features with an angle of 15˚ can be fabricated with an angle of 15.5˚. Typically, any section of the part that extends out at a 45˚ angle is printed with supports underneath to avoid warpage and collapsing during deposition. Currently, available open source or commercial slicing software can generate these supports automatically. In [5] many design rules for features joining or separating along the printing direction were given.

This limitation is important due to the disadvantages of adding supports such as extending the printing time and wasting raw filament.

In some cases, the overhanging parts are unavoidable and require necessary supports. The supports can be located in inaccessible regions and cannot be removed at the end. Such supports are fused using different material filaments that are usually water soluble which works if the solvent has access to the support. However, not all FDM printers have the capability to print using two extruders so that it can build the supports using different materials than the main part. Because of these disadvantages, it is important to know the limits of FDM printer to fabricate overhangs and bridges without supports.

As shown in Figure 6(b) an overhang part was printed to test the limits to print overhangs inclined with different angles. Using 0.20 mm layer thickness, three shells, 0.50 mm raster width, extrusion temperature 200˚C, and 40 mm/s printing speed with nozzle 0.5 mm on Taz 6^® FDM system using PLA filament of 3.0 mm. It was possible to print overhangs inclined with 60 degrees to 30 degrees from horizontal. Moreover, using the same settings, bridges extending by 10 mm to 40 mm were successfully fabricated without supports as shown in the Figure 6(c). More research effort is needed to optimize FDM for such features, as these limits can be further improved by optimizing the processing parameters.

In addition to overhangs, islands are another type of features that require a support during deposition. Islands, shown in Figure 7, are features that are not connected to the building bed and, for some time during the FDM process, are

isolated from the main body. However, if the part is fabricated in the opposite direction to what is illustrated in Figure 7, the need for the support will be eliminated as the island feature will have the flat top as a base. Thus, the orientation at which the part will be built might determine whether a part needs support or not.

FDM parts need a sufficient base area to ensure adhesion to the building bed. For example, spheres and curved sheets are very hard to fuse due to the small contact area with the building bed, which limits the possible orientations to build a given part with such features. In contrast, having a large base area increases the chances for warping.

Another feature that can be utilized using FDM is the ability to fuse around an insert objects. Although this feature is yet to be investigated thoroughly, in Kataria et al. [34] this feature was investigated for SLA and recently in Xiang “Anthony” Chen et al. [35] and Raymond R. Ma [36] work. It was shown that it can be applied for FDM as well.

In Klahna et al. [37] design guidelines for Snap-Fit Joints were given. The restrictions of AM were driven mainly from the anisotropy of the mechanical properties and the staircase effect on the surface of AM parts. The anisotropy may reduce the durability of load bearing features while the stair stepping hinders the movement of sliders, control rings and other kinematics. To incorporate these restrictions in the design, the build direction of the part at an early design phase should be determined carefully as in Figure 8.

In conclusion, during the design process the limited possible orientation for such features should be considered to facilitate the printing process. Similarly,

for some shapes one is restricted to specific fusing orientations such as for cylinders, cones, and pyramids. If a part has many features in which no optimized orientation is possible, dividing the part into printable parts might be the only option.

This term refers to how accurate the dimensions of the FDM part are in comparison to the input CAD drawing dimensions. The dimensional accuracy is affected by many parameters, mainly the X, Y, and Z positioning resolution, nozzle diameter and slicing thickness (layer thickness). The processing parameters during the printing process significantly affect the accuracy of the part as it can affect shrinkage, bonding and warping [38] - [42] . This is very important for assemblies. In a study for components benchmarking, it was found that 97.7% of the dimensions were in ±0.50 mm tolerance range and most of them were positive indicating a larger fabricated features than in the input CAD model [31] .

In a study done by the authors [17] , a total of 18 × 3 ASTM D683 specimens were fabricated using variant processing parameters. Each specimen was created in a set of three copies to check for the repeatability of the FDM process for same set of processing parameters.

The specimens were measured for their width, thickness and length. It was found that fabricating the part at different orientations affect the deviation in the dimensions. In addition, when the total dimension of the part was equal to an integer multiple of the layer thickness, a lower dimensional error is resulted.

For example, to create a solid cylindrical part with 25.4 mm (1 inch) height using FDM with a common nozzle diameter of 0.4 mm. The optimized cylindrical height to increase the dimensional accuracy will be either 25.5 mm or 25.2 mm instead of 25.4 mm if the slicing thickness used is 0.30 mm.

The dimensional accuracy is influenced by the processing parameters used during fabrication. To investigate the influence of some of these parameters in this work, a total of 8 tensile test specimens fabricated according to ASTM D638 type IV where created using different processing parameters. Four of the specimens, where fabricated using different extrusion temperatures and the rest of the processing parameters, were fixed. The remaining four samples were printed using different layer thicknesses and the rest of the processing parameters were the same for each sample. Each specimen was measured at the locations shown in Figure 9 and were compared to the dimensions of the input CAD model.

The measurements are averaged based on Table 2. The widths at locations W1, W2, and W3 were averaged as W and similarly for OWs and OTs, the dimensional error figures were generated as shown in Figure 10. From Figure 10(a) we can see that by increasing the extrusion temperature the dimensional error increases, which can be due to the uncontrolled flow of the extruded raster. Similarly, from Figure 10(b) reducing the layer thickness resulted in a reduction in the dimensional error parallel to that building direction, which is the thickness of the specimens.

This error can be avoided by selecting a layer thickness that is compatible with that dimension. For example, in Figure 10(b) we can see that the dimensional error at 0.10 mm and 0.25 mm layer height gave the smallest errors in the thickness of the tensile specimens. This occurred because both 0.10 mm and 0.25 mm can produce the thickness of the specimens using an integer number of layers, which is 35 layers for the 0.10 mm layer height and 14 layers for the 0.25 mm layer height. From Figure 10(c) it is indicated that the printing direction as described in Figure 9(a) affects the linear dimensional accuracy. This can be related to the variation between the raster width and the layer thickness in addition to the variation in the positioning accuracy of the nozzle between the x-y plan and the z direction.

During the design phase, clearances should be assigned between any meshing parts. For example, a pin should be fabricated with a slightly smaller diameter, while the hole where the pin is assembled to, should have a slightly larger inner diameter to facilitate the assembly. The Geneva mechanism was taken as a case study to investigate clearances for assembled parts in this work. To find the correct clearance firstly, a fitting calibration test was done as shown in Figure 11. Six small cylinders with diameters ranging from 4.00 mm to 1.50 mm in 0.50 increments were printed and 6 corresponding holes were printed. However, for the holes the samples were repeated with different assigned clearances ranging from 0.10 mm to 0.50 mm in 0.10 mm increments. It was found that the cylinders fit forcibly in the holes with 0.30 mm clearance and easily in the 0.40 mm while fitting loosely in the 0.50 mm. The 0.30 mm was used for the fixed assemblies in the Geneva mechanism such as the crank and its shaft, while the 0.40 mm was used for the moving joints such as the Geneva wheel and its driving pin, where the 5.0 mm slot was increased to 5.40 mm as shown in Figure 12.

FDM parts usually have high surface roughness and require surface finishing to get smooth surfaces that are comparable to the surface of injection molding parts. The surface roughness of a part is important in the case of FDM assemblies, especially for moving joints as the poor surface finish will affect the performance of the joint due to high friction. This can cause early part or assembly failure. The processing parameters of FDM directly affect the surface roughness of fabricated parts. FDM processing parameters highly influence the staircase effect such as the extrusion temperature, extrusion width, layer thickness and building orientation [43] - [48] . Typical values for surface roughness of a flat surface that does not suffer from staircase is Ra = 1.86 - 22.48 µm [49] . Whereas, for inclined surfaces suffering from staircase effect, the layer thickness and inclination angle play a great role [44] . It was shown that inclination angles that are

close to 30˚ have a higher surface roughness. In addition, increasing layer thickness, will increase the part’s surface roughness. A surface at 30˚ angle with layer thickness of 0.178 mm will have a surface roughness around 40 µm. If the surface finishing is an important aspect in the part, then the feasible building orientation will be a multi objective optimization problem. Another way to get a good surface finish is by choosing a chemically treatable filament such as ABS and post process the fused part [24] .

One of the advantages of AM technologies is that they eliminate the lead time needed until the production starts for mold design and production or setting up the production line. However, the fused process itself is very slow. Fusing time is stimulated by many factors such as the speed of the axis controlling the extruder, layer thickness and building orientation. Most commercial FDM printers have extruders that can move at speeds up to 200 mm/s or more. However, during extrusion a speed of 80 mm/s or lower is usually used to enhance the parts’ quality. By orienting the part’s largest surface area as parallel to the layer plane, the printing time can be minimized as shown in Figure 13. The estimated fabricating time for the same sample can be significantly reduced by orientation choice.

Fusing time is also driven by the layer thickness; the smaller the layer thickness is the longer the fusing time will be. In addition, avoiding the need for supports can reduce the fusing time significantly. That can be done either by eliminating unnecessary features that need support or by choosing an appropriate

building orientation that minimizes the need for support. In Teitelbaum et al. [50] many guidelines for choosing the appropriate building direction to minimize printing time were reported such as 1) build object with largest surface on bottom; 2) Maximize layer thickness; 3) Build holes facing upward; 4) Minimize number of overhangs; 5) Minimize the height by making the largest dimension parallel to the building bed.

Due to the nature of AM, parts suffer from high anisotropy in their mechanical properties. The designer should consider this anisotropy, which is highly affected by the processing parameters during the fusing. Therefore, it is important to consider anisotropy [51] [52] [53] [54] in the design process to optimize the functionality of the FDM produced parts [55] . To have strong parts, it is recommended to reduce the number of layers. That can be done by orienting the part flat [56] on its larger surface area as shown in Figure 13(b).

Increasing the layer thickness will yield a stronger part as well. It should be noted that FDM parts are usually stronger along the layer direction and weaker along the building direction, which is indicated by having a higher tensile strength and higher young’s modulus [17] . For that reason, the loading condition should be considered while designing the part to avoid and modify geometries that dictate building directions that might compromise the ability of the part to bear loads as desired.

Production cost is a primary factor that must be considered during the design stage. The cost of the FDM parts depends primarily on the material and the power consumed during fabrication. The material consumed can be divided into the part’s material and the construction material such as the material used for supports, rafts and brims which are for adhesion and warping reduction. The material needed for the part should be optimized by optimizing the part’s topology. The optimization takes into consideration the loading requirements and the part’s intended function. The construction material can be minimized by optimizing the building parameters such as building orientation which could reduce the need for supports. Constructional materials consumption can be reduced by using appropriately heated bed to minimize the need for rafts and brims. As a result, the building orientation of FDM parts is strongly correlated to cost [56] . On the other hand, the consumed power depends primarily on some processing parameters such as the nozzle diameter, the heated bed temperature and the time needed to finish the build.

Most FDM defects can be eliminated or reduced by using a good combination of processing parameters. However, some types of defects need to be considered during the design process as they might affect the performance of the part. The first defect is oozing or stringing, shown in Figure 14(a), which occurs when traces of small strings of polymer are left behind the nozzle when it is not extruding. This defect can be minimized by reducing the extrusion temperature and increasing the retraction of the filament during travel times. However, this defect can be persistent when fabricating features separated by small gaps. Strings caused by oozing can be mechanically removed or by chemical surface treatment for some materials. However, that can be challenging to accomplish for internal features and joints.

Another important defect is under extrusion, shown in Figure 14(b), which happens when the nozzle is not extruding enough plastic while filling the parts. Therefore, there are unintended gaps left between the contours and infill lines. This defect affects the part’s integrity and makes it weaker. In addition to rendering the defective parts not usable for fluid application since they are not water tight. This defect can be eliminated by increasing the extruded plastic, which can be done by various ways using commercial and open source slicers such as correct-

ing the extrusion multiplier or the entered filament diameter. Nevertheless, even after eliminating the under-extrusion defect, there is a gap in the literature about the pressures that FDM parts can handle compared to traditional manufactured parts. Some work presented an experimental methodology to investigate the main factors of profile errors in FDM parts [57] . A spiral shaped profile with cylinders with different radiuses where fused using different values of contour widths, contour depths, part raster widths and raster angles. It was found that contour width has the most obvious effect on both issues. The larger the width is, the better the profile accuracy is. On the other hand, a narrow width results in a dense and good appearance of the aperture area on the surface layer.