A Test Part for Evaluating the Accuracy and Resolution of a Polymer Powder Bed Fusion Process

Additive manufacturing (AM) offers a number of advantages over traditional subtractive machining techniques. AM technologies can expand the design space, allowing previously unachievable geometric complexity and yielding vastly improved part performance. For instance, aircraft manufacturers use AM to make innovative aircraft ductwork, which provides significant weight savings and reduces assembly time by consolidating parts [1]. AM can also improve sustainability of manufactured products, reduce time to market, and eliminate tooling costs which would be required for other manufacturing processes [2]. One factor preventing widespread adoption of AM in manufacturing is that designers have limited knowledge of the geometric capabilities and limitations of each process.

Metrology strategies are needed to characterize each AM process so that designers can compile detailed, statistical knowledge of the geometric resolution and accuracy for different features of interest. An uncharacterized process can lead to mistakes that require iteration to reach the final design intent, thereby diminishing the benefits of AM. A comprehensive metrology study can help designers reduce the number of design iterations and realize parts as they were originally intended in a more timely fashion. A critical component of a metrology study is a test part that incorporates the features of interest to the designer.

In addition to supporting statistical characterization of an AM process, a standard metrology test part can also help manufacturers monitor the performance of a specific machine over time. Machine operators can use test parts to evaluate machine performance compared with an “average” well-tuned machine and identify machines or process settings with poor geometric accuracy or resolution relative to their counterparts.

The focus of this paper is to present a compact test part and suggested metrology procedure for characterizing a polymer powder bed fusion (PBF) process. To support the primary goal of building a statistical database of geometric resolution and accuracy guidelines for designers, the test part must be comprehensive, containing a suite of commonly used design features. In addition, it must occupy a small build volume so that a large number of replicates can be fabricated efficiently, preferably by inserting the part into open spaces in existing builds. Its various features should also be relatively easy to measure.

Figure 1 illustrates several test parts from previous efforts to characterize AM processes [3–6]. A 2012 study by the authors investigated an extensive range of feature types including thin walls, holes, shaft clearance, and gaps for polymer PBF [5], and similar features are included in standard EOS test parts [7]. In these studies, each feature is investigated with a separate test part, leading to a low feature density that requires either multiple dedicated builds or large build volumes for each metrology study. Another AM test piece proposed by NIST in 2014 sought to generate a high feature density within a single part [6]. However, this part is not process-specific and fails to fully utilize the capabilities of polymer PBF, namely, the ability to build features without support structures. These studies highlight the need for a compact test part that encompasses a wide array of features that are of interest specifically for polymer PBF.

As shown in Fig. 2, the proposed metrology test part consists of five panels connected to a common base through small tabs. Each panel contains different geometric features and can be removed from the base for measurement. When built, the test part occupies a cube measuring 5 cm (2 in) on each side. Small protruded features beneath the base can be added to individually identify each cube. Table 1 describes the features embedded in the test part.¹

The first panel (Fig. 3) contains gaps varying between 5.0 mm and 12.5 mm in two directions. These gaps can be measured using calipers to determine the linear accuracy along a particular axis. Cylindricity can be evaluated from the three extruded circular features in the middle of the panel using a coordinate measuring machine with three-dimensional (3D) scanning capabilities. The scanner used in this study was a FARO ScanArm with a reported accuracy of ±0.025 mm. Cylindricity describes the feature's deviation from a perfect cylinder of the same diameter. A set of extruded and embossed hemispheres and cones can also be scanned, and their surfaces can be compared with the intended surface.

The second and third panels (Fig. 4) consist of gaps and holes as a function of wall thickness. The gaps and holes are measured at varying wall thicknesses because thicker walls raise the temperature of the surrounding powder bed, which can have the unintended effect of sintering surrounding powder [8]. This reduces the effective gap size and hole diameter in a phenomenon called oversintering. Based on the results from a previous study [6], the hole sizes are chosen to range from diameters that resolve easily to diameters that are too small to resolve, as a means of evaluating the resolution of the process. The accuracy of the gaps can be evaluated by measuring the gaps with calipers and comparing to the intended gap width. In this study, iGaging digital calipers with a resolution of ±0.01 mm are used. The accuracy of the holes can be evaluated by measuring them using pin gages or optical techniques. For hole measurements, a flatbed scanner and image processing matlab script are used after calibration to standard pin gages (±0.025 mm).

The fourth panel (Fig. 5) contains thin wall, thin rod, thick rod, and hinge features. Two different types of thin rods are included: rods with only one end fixed (unsupported) and rods with both ends fixed (supported). For the unsupported rods, aspect ratios (length/diameter) of 5 and 10 are used. For thin walls and rods, resolution is captured by visual inspection to determine whether the feature successfully formed. The accuracy of thin walls and rods can be evaluated by taking repeated measurements at different locations along the wall or rod with calipers and averaging the data. iGaging digital calipers with an accuracy of ±0.01 mm are used in this study. Two different hinge designs are included, and their resolution data are determined by checking whether the hinges can move freely. Three thick rods are included on the top of the part for evaluating dimensional accuracy using a 3D scanner.

The fifth panel (Fig. 6) contains features for lettering and mechanical snap-fits. Both raised and embossed lettering is included with varying font sizes and extrusion depths. Visual inspection using the evaluation criteria shown in Fig. 7 can determine the resolution of the lettering. Lettering that is clearly legible and contains no defects constitutes a “pass” rating. Lettering that is legible, but contains minor defects receives an “intermediate” rating. Lettering that is not legible or contains major defects is given a “fail” rating. For the snap-fits, there are pegs and slots of varying sizes. The slots are on a separate piece and can be used to test against each peg in order to determine which combination provides an acceptable fit. Slots that are either too small to fit around the peg or so large that the fit is loose are considered “failed” combinations. Peg and slot combinations that fit snugly without being easily separated are given a “pass” rating.

Finally, there are five small panels arranged in a fan connected to the base (Fig. 8). Each panel can be removed to gather surface roughness data. Surface roughness can be measured using either a mechanical or optical profilometer. For this study, a Zeta 3D optical profiler is used with a 10× objective lens. PBF is a layer-based process, causing a stair stepping effect when features are not oriented parallel or orthogonal to the build plane. The pieces are built at different angles relative to the base in order to measure how the layering affects surface roughness.

Since the primary goal of the proposed test part is to incorporate a variety of features of interest to polymer PBF part designers in a form that is compact enough to incorporate into open spaces in existing builds, it is important to compare the proposed test part with the test parts from previous metrology efforts. As shown in Table 2, feature density can be calculated by counting the total number of feature instances in the part and dividing the sum by the bounding volume of the test part. As shown in the table, the proposed test part's feature density is at least 4× greater than similar parts in the literature.

When planning a metrology study, it is important to build multiple copies of the test part under different conditions to characterize a PBF machine or process. Four factors in particular can affect the accuracy and resolution of a variety of features in a PBF process and can be easily modified and specified by the designer. The four factors are material choice, orientation of the test part within the build chamber, location of the test part within the build chamber, and machine identity. By varying each of these factors, a factorial experiment can be conducted.

Material selection is important for polymer PBF. It affects not only the mechanical and thermal properties of resulting parts but also sintering-related properties such as shrinkage, stress relaxation, oversintering, and surface roughness [9–11].

As a layer-based process, polymer PBF can contribute to anisotropic properties in finished parts. For this reason, the orientation in which features are positioned can dramatically affect the resulting feature quality [12,13].

Powder bed fusion for polymers requires a heated build chamber to alleviate residual stresses within the parts. Uneven heating of the build chamber can lead to thermal gradients across the powder surface. This variation within the build environment can have unexpected consequences such as varying accuracy and strength, depending on where the parts are built in the build chamber. The corners of the chamber tend to be cooler than the center, and parts located there tend to be less ductile with a higher degree of warping [14].

Machine identity can also impact part quality. Without considering specific machine tuning parameters, building the same part on different machines can yield different results due to differing machine calibration, repair and maintenance schedules, and other factors. Multiple replicates built on different machines allow machine variation to be taken into account [9].

It is possible to vary other process-specific variables such as laser power, scan speed, and hatch spacing, but these tuning variables are often fixed by the manufacturer when tuning each individual machine. Since designers typically do not adjust these variables, they may not be incorporated in a metrology study that is focused on generating design guidelines for a well-tuned machine. However, if the test part indicates poor resolution for a particular machine, it may be an indication that the machine needs to be tuned.

Selected results from two test cubes are documented in Table 3. Both cubes are built in the interior portion of the build chamber, on the same machine, and with the same material (a fire-retardant blend of Nylon 11—FR PA 11), but with different build orientations. Nominal dimensions correspond to the dimensions specified by the computer-aided design model. The measured wall thickness for both parts is larger than the nominal dimension, suggesting that thin walls tend to be oversized due to oversintering. The effects of different orientations are evident in the resolution of the thin walls. Walls built along the XY plane of the build chamber successfully resolved thicknesses as small as 0.3 mm, while YZ walls could only be resolved at thicknesses greater than 0.6 mm. The distinctions are likely due to XY resolution being governed by layer thickness, which is 100 μm in this case, versus YZ resolution being governed by scan patterns and spot size, which is 250 μm in this case.

The next step is to build multiple copies of the test part under a variety of conditions—material type, orientation, location, and machine of interest—in a factorial style experiment. Multiple copies are needed to build statistical distributions of the types of measurements reported in Table 3 and to support a statistical analysis, so that statistical design guidelines can be established. Statistical design guidelines can aid designers when dimensioning and tolerancing parts for polymer PBF, so that acceptable parts are built in the first iteration.

While the test part proposed here encompasses many different feature types, spatial limitations prevent some functional features from being included. Future studies could add functional features such as compliant structures, gears, and springs. Another opportunity for polymer PBF test parts would be to investigate various lattice structure designs. Strength and stiffness information could be gathered through mechanical testing, and geometric variation could be linked to performance variation. In addition, this test part assumes that the machines are well-tuned for manufacturing functional parts and focuses on the effect of designer-specified parameters (orientation and material) on accuracy and resolution. Other process-specific test parts and process monitoring technologies may be useful for manufacturers to tune their machines and associated manufacturer-specified process parameters to improve resolution and accuracy, along with part strength and other outcomes.

Each AM technology has its own unique capabilities and challenges specific to the process. Consequently, a “one-size-fits-all” universal test part is unable to effectively describe each process. The test part proposed here is specific to polymer PBF and would not be well-suited for other technologies. Even metal PBF would be unsuitable because of the need for support structures. Customized test parts for each AM process could provide designers with the information needed not only to select the appropriate process for each design but also to utilize the process to its fullest extent.

The authors gratefully acknowledge the support from Stratasys Direct Manufacturing, especially Mr. Steven Kubiak. This effort was performed through the National Center for Defense Manufacturing and Machining under the America Makes Program entitled “A Design Guidance System for Additive Manufacturing (Project 4053)”. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of Air Force Research Laboratory or the U.S. Government.

• Air Force Research Laboratory (Agreement Nos. FA8650-12-2-7230 and FA8650-16-2-5700).