Additive Manufacturing (AM) is the term used for technologies that produce three-dimensional (3D) functional parts from nominal computer-assisted design (CAD) files using typically layer-by-layer material deposition techniques. These technologies do not require conventional tooling to build components since the shape is produced by adding, rather than removing or deforming, material. The material can be polymer, metal, composite, ceramic, concrete or even human cells. Many AM processes have been developed and are commercially available, including Stereolithography (STL), Fused Deposition Modelling (FDM), Three-Dimensional Printing (3DP), Powder Bed Fusion (PBF), Direct Metal Deposition (DMD) and Sheet Lamination (SL). The PBF technologies include two variants depending on the nature of the heat source: the Electron Beam Powder Bed Fusion (EBPBF) and the Laser Powder Bed Fusion (LPBF). AM standard terminologies are framed by ISO/ASTM 52900:2015 [1
] and their general principles are described in ISO/ASTM52901-16 [2
LPBF is one of the most used processes for metallic AM, which builds three-dimensional parts directly from metal powder. In a chamber filled with inert gas, a high-power laser beam selectively scans a thin layer of metallic powder, resulting in local melting. Dimensional accuracy prediction and control remains a major concern when it comes to the industrial adoption of this technology.
To investigate the process feasibility and transpose the information collected into the process performance, many researchers have proposed AM test artifacts that are meant to quantify the capabilities, limitations and accuracy of the machine and the process, and to diagnose specific processing defects [3
]. Richter and Jacobs [4
] suggested that the standard test artifact should be large enough to test the performance of the machine near the edges of the platform as well as near the center, have a substantial number of small, medium and large features and have both holes and bosses to aid in verifying beam width compensation. In addition, it should not take too long to build, nor consume a large quantity of material, and should be easy to measure. Byun et al. [5
] added that the test part should also have evaluation features to assess whether or not it is possible to manufacture fine features under the specific AM process. Accordingly, these fine features should be set up in all axes, and their size should be varied while considering the improvement in the process mechanisms and resolutions of AM machines. Kruth et al. [6
] stated that such artifacts should not only be used to analyse the process limitations, but also to optimise the process iteratively. Some other authors have made notable advances in the field of AM artifact design [7
Globally, to date, researchers have focused more on the feasibility issues and less on the statistical effects of processing and post-processing operations on final part deviations. Knowing that geometrical deviations strongly impact the service performance of structural parts, this lack of reliable metrological information hinders widespread industrial adoption of AM technologies.
To fill this gap, different authors have studied and tried to predict dimensional and geometrical deviations of AM parts with varying levels of success. For example, Singh et al. [18
] quantified the effects of laser power, scan spacing, powder bed temperature, hatch length and scan count on the geometrical deviations of Selective Laser Sintering (SLS) polyamide parts. Next, Huang et al. [19
] investigated the compensation of the geometrical deviations on SLA SI500 (resin) parts. They presented a statistical predictive compensation approach to predict and improve the quality of cylindrical and prismatic parts. However, only the XY plane deformation errors were taken into account, while the Z coordinate was ignored. Moreover, after the compensation, the parts still presented the same systematic deformation pattern as before, and only the average profile deviation was improved, but not entirely corrected.
Similar works can be found on metal AM. For example, Zongo et al. [20
] carried out an intra- and inter-repeatability study of profile deviations, and the results of the investigation demonstrated that the LPBF performance for geometrical variations of 147 identical AlSi10Mg parts falls within a range of 230 µm at a 99.73% confidence level. Using the same process, powder and stress relief annealing heat treatment, Calignano et al. [21
] studied the dimensional limits of geometries with sharp edges. They quantified the effect of STL file tolerances on the printed part deviations. Next, Van Bael et al. [22
] investigated the geometrical controllability of LPBF-built Ti6Al4V porous structures by means of a feedback loop between the design and the printed part deviations. After two iterations, the average pore size mismatch was decreased from 45 to 5%. Li et al. [23
] investigated the effect of micro-vibrations on the final part’s porosity and mechanical properties. They demonstrated that 969 Hz vibrations can decrease the density of AlSi10Mg parts printed using a KUKA six-axis robot from 100 to 99.1%. Next, Liu et al. [24
] demonstrated that if each layer is scanned twice, the part density increases by 0.1%, and if it is scanned three times, the density is increased further by 0.3%. It is evident that these modifications in processing sequence must have affected not only the part density but also its geometry; this last aspect was not considered in these works.
Notwithstanding the above, the number of metrological studies of AM processes is still limited. This paper isolates and quantifies the intra- (same part) and inter- (different parts) scale effect, and the material concentration, stress relief (SR), part removal (PR) and micro shot peening effects on the 3D profile deviations of selected simplified test components. The results of these analyses can serve as an accessible experimental database for the validation of numerical models intended to predict the geometrical and dimensional deviations of LPBF parts. The ultimate goal of this study is to improve the design support for the LPBF technologies.
The paper is organised as follows. Section 2
describes the part used and the experiment protocol followed. The results are presented and discussed in Section 3
. Finally, a summary is provided and future work is described in Section 4
To isolate, test and quantify the effects influencing LPBF part geometrical deviations, two test artifacts were designed. The first, referred to as the Shape A artifact, is a cylindrical part containing a coaxial cylindrical pocket displaying a diameter of half of the full cylinder (Figure 1
). This artifact was manufactured in 12 versions. The variables are the pocket depth (from P1 being one-fourth depth to P4 completely hollow) and the part size (from S1 small to S3 large), as described in Figure 1
. The Shape B artifact is a four-step cylindrical pyramid. This artifact was manufactured in six versions. The variables are the absence or presence of a coaxial pocket with a diameter of half of the smallest cylinder (P0 means no pocket and P4 stands for completely hollow) and the part’s relative scale S1–S3, as described in Figure 1
The pocket depth variant (P) allows for the isolation and quantification of the impact of the material concentration on the part deviations, while the scale variant (S) allows for the isolation and quantification of the impact of the scale effect on the part deviations.
All of the parts of this study were printed using an EOSINT M280 system (Electro Optical System, Germany), and AlSi10Mg powder with the AlSi10Mg_Speed 103 process parameter set (laser power 370 W, scanning speed 1300 mm/s, hatching space 0.19 mm and layer thickness 30 µm). The used AlSi10Mg powder has a particle size distribution of D10 = 12.8 µm, D50 = 27.7 µm and D90 = 51.3 µm; the tap density 1.358 g/cm3
; the apparent density 1.081 g/cm3
and the Hausner ratio 1.256 [25
]. Prior to printing the test artifacts, a beam offset calibration was performed. To this end, the calibration part was printed and inspected (Figure 2
a). The part was then scanned by means of a Renishaw PH10 probe mounted on a Mitutoyo Coordinate Measuring Machine (CMM) (4 µm accuracy at a 95% confidence level). Data were collected on the as-built part and after stress relief of 300 °C for two hours. The results of the as-built part’s inspection were used to assess the quality of the beam offset correction protocol suggested by the LPBF system manufacturer.
After the beam offset correction, eighteen (18) AlSi10Mg artifacts were printed using the same LPBF system, material and parameters for the calibration part (Figure 2
b). The point cloud of the parts printed was obtained by means of a Metris LC50 laser scan mounted on a CMM (
7 µm accuracy at a 95% confidence level) (Figure 2
). Before each scan, the devices were calibrated using a master sphere and data collection was performed under multiple angles to maximize the information collection on inner surfaces. A real-time visualisation was possible with the Focus Inspector, a specialised software application. A thin layer of talcum powder was used to reduce part surface reflection. As a result, the potential point cloud density was increased to ensure the best measurements. The point clouds were then assembled from different angles and cleaned. Following a first Geometrical Deviations Extraction (GDE), a second GDE was carried out after stress relief annealing at 300 °C for two hours, as recommended by the LPBF system manufacturer. The heat treatment was conducted under an argon atmosphere followed by air-forced cooling down to room temperature. Next, a third GDE was carried out after the parts were removed from the plate using a 2°mm thick saw on a horizontal setup. Finally, as suggested by the LBPF system manufacturer, micro shot peening was then applied using an IEPCO MICRO 750 S system with the IEPCONORM-A agent (0.2–0.4 mm grain size of crushed corncob), applied with a 3 bar pressure at the perpendicular angle to the specimen surface, at a 3–5 cm distance, before the fourth and final GDE. The Gaussian best-fit technique and data alignment were performed on the GDEs using PolyWorks® v.16 (Innovmetric Metrological Software). The data were then loaded into Matlab® 2017b (MathWorks), using a code to extract the deviation at each point. Minitab® v.17 (statistical software by Minitab Inc.) was used for the graphic and statistical studies.
To sum up, the calibration analysis (Analysis A) and three types of deviation analyses were performed based on the ASME Y14.5 (2009) tolerancing standard: The scale effect analysis (Analysis B), the part material concentration effect analysis (Analysis C) and the post-processing effect analysis (Analysis D). Each of the deviation analyses was carried out using Shapes A and B artifacts.
2.1. Analysis A—Calibration Part
The calibration part inspection was conducted according to the plan provided by EOS. The plan specifies which feature needs to be inspected (see Figure 4a), and how to use the provided Excel sheet to calculate the beam offset, using the results of such an inspection. This analysis was carried out using the part inspection results. The correlation between the feature inspected nominal size and the stress relief effect was investigated.
2.2. Analysis B—Scale Effect
The scale effect analysis was carried out using Shapes A and B. First, an intra-part scale effect (scale effect on different features of the same printed part) was carried out using four cylindrical features of the Shape B artifacts presented in Figure 3
b. The framed symbols A and B on the shapes represent the features used for the datum alignment.
An inter-part scale effect study (comparing the same shape with different part sizes) was carried out using Shapes A and B. Regarding the Shape A artifact, the as-built external diameter deviations were extracted and compared for three different sizes, S1, S2 and S3 (Figure 3
a). For the Shape B artifact, the 3D profile deviation nonparametric cumulative distribution functions (CDFs) of all the parts at S1 and S2, as well as of two parts with no pocket (P0) at S3, were used for this study (Figure 3
2.3. Analysis C—Material Concentration Effect
In topology optimisation, the material concentration or “pseudo-density” factor describes the layout of the material in a part within a given design space [22
]. In this study, the material concentration effect was studied by means of the pockets variant (P0–P4) (Figure 1
). To this end, the A-S3-P0 and A-S3-P4 artifacts were printed twice on the plate to confirm the material concentration effect on each of the two parts with different pocket depths. Comparisons were made between parts with the same global shapes and sizes, but pocketed at different depths. Regarding the Shape A artifact, the material concentration effect was studied using the 3D profile deviation of artifacts having four different pockets depths (P1–P4). The results are presented at each post-processing step (PPS). The parts used are at their biggest scale (S3). With the Shape B artifact, the part material concentration effect was studied using the 3D profile deviations of two P0 and P4 pockets depths, with each variation being printed twice. The mean 3D profile deviation is presented at each PPS. The parts used are at their biggest scale (S3).
2.4. Analysis D—Post-Processing Effect
The post-processing effect was studied by observing overall 3D profile deviations and specific feature evolution before (as-built (I)) and after each post-processing step (PPS): stress relief (II), part removal (III), and micro shot peening (IV). The analysis was carried out using Shapes A and B. Regarding the Shape A artifact, the overall 3D profile deviation mean values, before and after stress relief, and the deviation of the external diameter at each PPS were extracted and are presented. The nominal values are specified in Figure 3
a. For the Shape B artifact, the overall 3D profile deviation mean values of four parts at S3 were determined. Furthermore, the diameter deviations of four cylinders of the pyramidal shape (Figure 3
b) were extracted on each Shape B artifact, at each PPS.
General note: By definition, the deviation is the difference between the nominal value (as defined in the CAD file) and the experimental value extracted from the measurements. Positive deviation means that the measured value is greater than the nominal in the Maximum Material Condition (MMC) direction. Negative deviation means that the measured value is smaller than the nominal in the Least Material Condition (LMC) direction, as in ASME Y14.5.1 . The interval bars presented on the graphics are the measurement system uncertainty at a 95% confidence level.
This work is designed to isolate, evaluate and quantify the metrological performances of the laser powder bed fusion of metal powders. It presents and discusses an analysis performed on 18 specially designed artifacts printed using AlSi10Mg powder and an EOS M280 LPBF system. The study quantifies the intra- and inter- scale effects, the part material concentration effects, the printing deviations behaviour (I), the stress relief effect (II), the part removal effect (II) and the micro shot peening effect (IV) on the 3D profile deviations of the parts. Based on the obtained results, the following conclusions can be drawn:
There exists an intra-part scale effect; different features of the same part having different nominal sizes manifest different deviations.
For the parts of the same size, the lower the material concentration, the lower the number of observed deviations.
Stress relief heat treatment reduces the intra- and inter-part scale effects by expanding (MMC) the larger features more than the smaller features.
The parts removal operation globally increases the parts deviations.
Micro shot peening has a positive effect on the surface roughness, but systematically reduces the parts size.
The results of this study will serve as an accessible database of experimental values carried out according to the GD&T ASME Y14.5 (2009) standard. Therefore, they can be used to validate numerical models that aim to predict the geometrical and dimensional deviations of parts manufactured using the same process and machine, with the same powder. Enhancing the efficiency of the design for AM (DFAM) numerical models will have a positive impact on the competitiveness of AM and could boost its adoption by high-technology industries where the production cycle would greatly benefit from these technologies.