Multiscale Modeling Framework for Defect Generation in Metal Powder Bed Fusion Process to Correlate Process Parameters and Structural Properties

Abstract

Powder Bed Fusion (PBF) is one of the most popular additive manufacturing methods employed extensively to fabricate complex parts especially in industries with stringent standard criteria, including aerospace, medical, and defense. DMLS/PBF fabrication of parts that is free of defects represents major challenges. A comprehensive study of thermal defects, contributing parameters, and their correlation is necessary to better understand how process specifications initiate these defects. Monitoring & controlling temperature and its distribution throughout a layer under fabrication is an effective and efficient proxy to controlling process thermal evolution, which is a completely experimental technique. This being highly costly specifically for metal printing, computer-based numerical simulation can significantly help the identification of temperature distribution during the printing process. In this paper, a multiscale modeling technique is demonstrated with commercially available software tools to correlate the defect generation in metal PBF process and significant process parameters. This technique can help efficiently design the process setting in addition to or even absence of experimental monitoring data. This research work is a part of a larger project of closed-loop control strategy development using physics-based modeling and graph-based artificial neural network implementation for reducing thermally induced part defects in metal 3D printed process.

1. Introduction

Additive Manufacturing (AM), also known as Rapid Prototyping and 3D Printing is a three-dimensional fabrication process, executed by adding materials in layers. This is a revolutionary product development and manufacturing method, especially in the age when we are experiencing a mew industrial revolution. Small, relatively simple products may only make use of AM for visualization models, while larger, more complex products with greater engineering content may involve AM during numerous stages and iterations throughout the development process [1]. Among many different classes of AM processes, Direct Metal Laser Sintering (DMLS) is a widely used metal part manufacturing method. This process is carried on by using laser power to melt powder metal particles leading to a complete print of the part using the desired 3D solid CAD model data by a layer-by-layer process [2]. The key is to melt the material in a controlled fashion without creating a high accumulation of heat, so that when the energy source is removed, the molten material rapidly solidifies again. Recent involvement of large companies in developing metal AM processes have opened up the market significantly. As a result, machine accuracy, speed, cost, and quality of production have become apparent crucial factors in metal 3D printing nowadays. High quality metal parts with complex geometry can be produced in PBF process. At the same time the manufacturer needs to ensure part-quality, consistency, and competitive pricing to run and sustain a successful metal printing business. Here comes the requirement for deeper understanding of the process so that each and every step can be improved, optimized and controlled for higher efficiency, reliability, and production quality. The metal printing technology is in practice for more than a decade now, yet researchers are working till date to understand the physics of powder bed fusion process, and this is an ongoing pursuit. In this paper, a multiscale modeling technique is described which aims at correlating the PBF process parameter and their impact on single layer as well as subsequent printing defects on build part.

2. Background

Powder bed Fusion (PBF) or Direct metal laser sintering (DMLS) is an additive manufacturing process to build fully functional rapid metal prototypes and production parts. A selective sintering process is carried on using laser power to melt powder metal particles leading to a complete print of the part by a layer-by-layer process [1]. DMLS can be used to create parts from a variety of metal alloys. This makes it more popular than many other 3D printing techniques that are designed to work with specific metal alloys or polymer-based materials. One big advantage of DMLS is that custom manufactured parts made using this process tend to be free of internal defects and residual stress that typically occur in parts that have been made using more traditional manufacturing methods. This ability to create defect-free parts is critical when parts are to be used in a high-stress environment such as the automotive or aerospace industries.

2.1 Defects in PBF build parts

Several common defects are encountered in metal PBF processes, which lead to weakening of mechanical properties of the build part. These defects occur during fabrication and/or post -processing operation. The presence of defects limits the industry-wide spread of this technology as a result of insufficient repeatability, reliability, and precision. Such defects can be classified and analyzed based on how they affect the printed part, how they generate, how common and significant they are for the overall quality of the build part etc. In addition to that, different process parameters influence the generation and propagation of different anomalies in the print. Many studies have been conducted to identify the defects in AM processes [3, 4, 5].

In a recent work [6], the common defects in additive manufacturing have been classified and reviewed on the basis of geometry, surface quality, microstructure, and mechanical properties. For example, on the basis of geometry and dimension, there can be (a) geometry inaccuracy (form deviation) and (b) dimension inaccuracy (size deviation). Again, common surface quality related defects are surface roughness, balling, surface oxidation etc., while anisotropy, heterogeneity, porosity etc. are microstructure related defects found in 3D printed builds. A comprehensive classification is shown in Figure 1 [6].

Figure 1.

Defects in metal PBF.

2.2 How do process parameters affect printing defects

The significant process parameters in the metal powder bed fusion (also known as selective laser sintering process) such as laser power, scanning speed, hatching distance, scanning strategy etc. affect the generation of printing defects. It is crucial to identify the combination of these parameters to obtain the required level of quality of the product. For instance, higher laser power increases the thermal shrinkage, and higher scanning speed hatch spacing lowers the thermal shrinkage [7]. The laser power and scanning strategy contributed to the temperature variation, that leads to non-uniform shrinkage in a particular layer [8]. Part weight, build chamber temperature, cooling rate, layer thickness and material can affect thermal shrinkage in a way that shrinkage decreases with increasing layer thickness, part bed temperature, and interval time [9, 10, 11].

Surface roughness depends on numerous parameters, some of which can be controlled, while few others are not controllable from the designer and operator’s perspective. For lower scan speeds, the average roughness decreases with increase in speed, while in the higher speed range, it remains constant [11, 12, 13, 14, 15, 16]. Moreover, warping and distortion that impact the surface quality, are highly dependent on thermal phenomena during the printing process. In addition to laser power and scanning speed, the thermal gradient between scanning zones can impact the quality of each fabricated layer. Surface geometry and fundamental geometric features of orientation, thickness etc. also have impact of geometric errors and surface quality [17]. Smaller hatch spacing seems to be beneficial in this context as it induces gradual temperature increase of powder bed and slower cooling rate. In addition to this, the initial powder spreading in LPBF also influences the layer quality, and thus consequentially impacts the porosity of produced parts [18].

Typical manufactured components using traditional manufacturing methods (milling, drilling, surface grinding etc.) comes from a solid building block of material and they are not porous unless porosity is induced by design. But in additive manufacturing, porosity of the build parts is a very common occurrence. Poor wetting, powder packing density, gas flow condition, entrapped gas etc. play vital roles behind the unwanted porosity of printed parts.

Among the numerous process parameters that contribute to defect generation, most vital ones are laser power, laser scanning speed, laser spot size, powder size, layer thickness, external pressure and material’s absorptivity. It is important to understand how the process parameters impact the generation of defects and what are the signatures that relate to the defects. The process-structure–property (PSP) relationships have been under discussion and research works are published earlier as well [19, 20, 21]. In recent years, use of machine learning and smart manufacturing has started to be used in AM field [22, 23, 24, 25, 26, 27, 28, 29, 30, 31], which also provides inspiration to use that for our research goal.

3. Generation of defects

The two major phenomena that contributes to the generation of defects in powder bed fusion are Keyhole and Spattering. In this section, these modes are discussed to understand the underlying physics of why and how a defect may occur in a layer or whole printed part. In a typical laser processing case, two modes of heat transfer participate in defect generation: namely, conduction mode and keyhole mode [32]. Conduction is melting is controlled by heat conduction. In keyhole mode, input power density is sufficient or high enough to vaporize the metal powder. It then creates a much deeper hole or cavity compared to conduction mode. Collapse of such cavities can leave voids in printed parts [33], hence, conduction mode of heat transfer is desired in laser additive manufacturing.

3.1 Keyhole mode

As mentioned earlier, keyhole is a major cause of forming pores and voids. Basic criterion of identifying the keyhole mode is shown in Eq. (1)

$$\frac{2\mathrm{d}}{\mathrm{w}}\ge 1$$E1

where d is the remelted depth, w is the melt pool width, and the quantity, $d/w$, is the normalized depth [34, 35, 36]. There are few other empirical formulae and methods proposed by researchers [37]. These enable the designers to identify the preferred conduction mode in L-PBF conveniently. By experimental methods, it was found that normalized depth is proportionally related to product of power density and square root of laser interaction time [38]. It obeys Eq. (2) as following:

$$\frac{d}{w}\propto \frac{P}{\pi {\sigma }^2}\times \sqrt{\frac{2\sigma }{\upsilon }}$$E2

where P is laser power, σ is laser spot size, and ν is laser scan speed.

3.2 Spattering phenomenon

Spattering is a physical phenomenon in laser powder bed fusion process observed from experiments [39, 40]. It is considered to be the major cause of the structural defects in the printed products. This is a complex physical phenomenon that requires experimental procedures to observe and challenging to model properly in a FEA computer numerical analysis.

From previous experimental tests, it is found that some spatters have a propensity to merge together and form larger particles. Through their comparison study between multi-laser and single laser scanning, Andani et al. [41] found that high number of working lasers can induce higher recoil pressure above melting pool and as a result more spatters are ejected for molten pool. In case of stationary laser impulse, as time goes by, vaporization occurs after melting and generates intensive vapor jet that ejects metal powders. With the surrounding pressure of inert gas decreasing, it forces the particles surrounding the molten pool to move forward. In this manner, metal powders are ejected with a large divergence angle as vapor can expand freely. This physical event can be observed as “Spattering” by using high-speed X-ray monitoring. Figure 2 [42] shows a schematic of pressure and time dependent spattering mechanism.

Figure 2.

Pressure and time dependent spattering mechanism.

4. Multiscale modeling technique

As mentioned earlier, spattering mechanism is difficult to model using traditional simulation approach due to its multi-phase nature and complexity in physics. This applies to the overall additive manufacturing or metal printing process as well. Molecular Dynamics (MD) can be an option in this area, considering only the atomic motion without any other information such as thermal conductivity or specific heat of the material. But, due to its nanoscopic nature, available computational capacity of today falls way behind the complexity of any feasible MD model. Some experimental techniques can be used as demonstrate in [43] to understand he metallurgical defects in PBF parts. But investigate generation of defects due to the process itself is significantly difficult. Computational Fluid Dynamics (CFD) modeling can be applied to simulate sintering process involved in L-PBF, as this can demonstrate the molten pool dynamics. But the fact that solid metal powders can only be stationary in CFD models, can limit the scope of complete simulation of AM process to understand the process-structure–property relationships. On the other hand, a simple meso-scale simulation using finite element analysis can help to correlate the build part defects (such as, deformation, buckling, holes, staircase effect etc.) with AM process parameters. This in turn initiates the idea of a multi-scale modeling scheme for overall understanding of the AM process.

A CFD modeling tool can correlate the process parameters (laser power, scanner speed etc.) to the defect generation mode or phenomenon (e.g., keyhole and spattering). Finite element simulation can predict temperature distribution over each layer with appropriate beam size and diameter, using material properties of the metal powder and subsequent build. The complexity of simulation depends on the available physics on the modeling scheme. However, these computational studies and any other experimental observations are not sufficient enough to create a comprehensive design and optimization method. To obtain a generalized process parameter optimization technique, a surrogate model is desirable to alleviate the overbearing requirements of frequent experiments/ simulations.

An alternative solution of this impasse is application of machine learning (ML) model. A supervised machine learning algorithm would be able to create an artificial neural network intelligent enough to predict the defect generation and hence recommend suitable combination of process parameters to generate flawless/ near- flawless printed parts. But performance of such algorithm depends on a well training program with appropriate input–output data. This data would come from numerical simulation and/or experimental testing. The multiscale modeling with different level of simulation will add to the experimental data that can be obtained in a real-time laser powder bed setting.

This multiscale modeling work is divided into three major steps:

• Micro-scale Computational Fluid Dynamics Analysis (additive metal powder size varies in the range of 50–100 μm. The simulation covers powder bed preparation, deposition and spreading and melt pool analysis in micro-scale).

• Meso scale Finite Element Analysis (focuses on macroscale geometry, stress, strain, deformation and build time).

• Validation study.

4.1 CFD analysis

A commercially available CFD software named FLOW-3D specializing in 3D transient flows with free surfaces is used in this study. It follows Volume of Fluid (VoF) method and TruVOF algorithm. The physics behind general welding and laser melting in PBF is similar, hence it is used in the CFD modeling scheme. Key factors involved here are laser beam motion, shield gas pressure, laser heat flux profile distribution, evaporation pressure and multiple laser reflections.

With the input of material properties of metal powder, powder size, bed size, the first step of the simulation is powder bed preparation. After the bed spreading is simulated, we can input the process parameters such as laser power, beam diameter etc. for the designed geometry and it will subsequently complete the laser melting simulation. Figure 3 the laser melting of metal and subsequent melt pool formation with temperature distribution [44]. Keyhole-induced porosity formation is observed in Figure 4 [45].

Figure 3.

Laser keyhole welding modeling using CFD.

Figure 4.

Keyhole induced porosity formation in L-PBF process.

Understanding the evolution of melt pool depending on varying process parameters can provide information on temperature distribution and porosity formation as shown in Figure 5 [44]. This is a crucial information for the multiscale-surrogate model. Rise in recoil pressure at the bottom of keyhole and increased surface tension at the upper region generates an irregular pore. The pore is then pushed to the back of melt pool by string downward flow and then it gets trapped by the advancing solidification front. Using this aforementioned CFD model, we can accurately represent the fluid flow at the melt pool at 1–10 μm length scale [44]. It also demonstrates defective design space and melt pool geometries, predicts compositional segregation and phase nucleation and growth.

Figure 5.

Melt pool formation in L-PBF, (a) 3D view, (b) 2D view, (c) sectional view.

4.2 FEA model

A recently developed and launched suite in ANSYS workbench called “ANSYS Additive” provides capabilities to simulate the complete metal additive printing process in L-PBF method. In addition to the common uses of creating and optimizing design solutions for AM purposes, it can also be used for understanding the metallurgical properties of the printed parts with porosity and microstructure prediction. Accessible data throughout the process ensures the traceability and hence enhances the feasibility of a parametric study using DOE (Design of experiments) technique.

In this method, we incorporate the layer-wise structure development and time discretization (exposure time of single layers is in milliseconds, but total build time can be hours) by using a detailed model for single layer and global model for whole structure (Figure 6). This is called lumped layer approach. The details of the theoretic development of this model can be found in ANSYS Additive training website [46].

Figure 6.

Lumped layer approach; top: Single layer detailed model, bottom: Whole structure global model.

Necessary material information to perform this simulation are density, thermal conductivity, heat capacity, structural properties such as Young’s modulus, Poisson’s ratios, thermal expansion co-efficient, and stress–strain data. Material properties of common AM metals can be obtained from pre-existing database of Workbench Additive and/or modified with specific case study or experimentally driven data.

Using this technique, it is possible to simulate the building of whole structure, without the complexity of looking into the details of each layer formation and related void generations. As we already know the parametric relationships between defect generation and process parameters from the aforesaid CFD model in Section 5.1, this FEA model complements it with the whole structural information of the printing process.

The main challenge for the FEA simulation is local discretization, by which we mean the dimensions of the laser spot are in μm, whereas the dimension of the whole structure is in cm. Moreover, exposure to a single layer occurs in ms (milli second), but the full built takes longer time, even couple of hours. This issue can be taken care of by applying a lumped layer approach in ANSYS additive. A quick simulation catches global stress/strain as well as the distortion that takes place during printing. The whole structure is simulated in a global simulation model, where several subsequent simulations are done with status and boundary conditions are updated consequentially. For such simulations, a few assumptions are made. These are:

• thermal and structural physics are uncoupled,

• no use of laser beam (consider whole layer at a time),

• several powder layers are lumped into one FE layer, and

• no melted powder is modeled here [46].

This is acceptable as we are focusing on the melt pool in the CFD simulation, and the combined modeling scheme will provide a comprehensive identification of all the relevant phenomena in the printing process.

A sequentially weakly coupled transient thermal-structural analysis is performed in this FEA model. Similar to conventional structural or thermal analysis in ANSYS workbench, we can import any kind of CAD file to this additive suite. Next steps include body cartesian mesh generation, creating named selections for build, support, and base. The simulation wizard is capable of automatically generating the support and base, so the designer only needs to consider the design of build part. After that, the build settings should be defined. This includes machine setting, deposition thickness, hatch spacing, laser speed, time between layers etc. Thermal boundary conditions include preheat temperature, gas/powder temperature, convection coefficients, and cooldown temperature (usually room temperature). Common deposit thickness is 0.001 mm–0.1 mm. A very large model will need higher computation power and time, in such cases, High performance computing (HPC) will help.

Using result tracker for temperature and displacement, the user can control the progress of print process during the solution as shown in Figure 7 [46]. Moreover, it is possible to switch between automatic and manual mode for result tracker.

Figure 7.

Result tracking during solution in FEA model, top: Tracking global maximum temperature, bottom: Build progression shown from left to right.

As mentioned earlier it is a weakly coupled FEA analysis, a transient thermal analysis is first conducted with appropriate properties and boundary conditions. Then the results are fed into a structural analysis model, and we would finally obtain the deformation, bucking etc. on the whole structure. A general model tree is shown in Figure 8 for easy understanding of reader.

Figure 8.

Typical model tree of coupled AM FEA study.

4.3 Validation

To develop a reliable simulation model and build a dependable training data set, it is necessary to validate the model and its input–output information. This requires experimental testing of the designed case studies in a real-time metal additive printing setup. There are several small-scale additive metal printers are available in the market which are cost-effective, but ideally, they do not reciprocate all the complexities of a full-size metal printer. Hence, we will perform experimental testing on a real time metal PBF setup to validate the multiscale numerical model.

These experiments will serve two purposes:

• Provide thermography data for building a training data set.

• Validate simulation model by replacing few user cases with specified process parameter combinations.

Experiments will be conducted in an open-structure metal AM printer, PANDA 11 (Figure 9). This open structure allows the system to monitor the build system and track temperature data for each layer. This system also enables closed-loop feedback operated online monitoring and control system. This is a part of the ongoing research that the authors are working currently. Details of the project can be found in recent publication of this research group [1]. This project involves understanding the physics of defect generation in metal powder bed fusion and using machine learning (ML) algorithm to implement the knowledge in automated process parameter selection to minimize printing defects. The machine learning algorithm under consideration is a graph based spatio-temporal convolutional neural network (ST-GCN) that will be trained using the results obtained in CFD and FEA modeling. The code will also incorporate genetic algorithm and/or game theoretic model to optimize the process parameters in order to minimize the defect generation. Once the ML code is trained and tested, it will be implemented using the device driver of the Open additive machine (shown in Figure 9). During operation, an online monitoring system using IR camera will be used to track the thermal history. This spatio-temporal temperature data will work as input to the ML algorithm, and finally using optimization theory, the device driver will receive information on optimized combination of process parameters, that will be activated for the next layer of printing. This is a novel idea for controlling the defect generation in metal additive manufacturing using process parameters and physics-based understanding of the process.

Figure 9.

Open architecture 3D metal printer PANDA.

5. Data collection

The information obtained from FEA study and validation study can be used in order to design and develop a spatio-temporal neural network. This artificial intelligence algorithm can provide optimized process parameters to produce parts with minimum defects. The experimental procedure being costly and designed to be at the end of the project, we started with initial FEA study at CFD level. We have created a Design of Experiment (DOE) study using CFD analysis. We have used two different material powder, three different laser power setting and three different scanning speed settings to perform the DOE. Moreover, we are considering single pass single layer melting only. The main purpose of this DOE is to investigate the diverse data obtainable from the simulations and understand impact of process parameter on the temperature map on PBF processes. So, there are 2 x 3 x 3 = 18 cases in design of experiments using full factorial method. Table 1 shows the DOE setup.

Case number	Material	Laser power (W)	Scanning speed (cm/s)	Case number	Material	Laser power (W)	Scanning speed (cm/s)
1.1	Inconel718	280	220	2.1	PH-316 SS	280	220
1.2	Inconel718	280	300	2.2	PH-316 SS	280	300
1.3	Inconel718	280	350	2.3	PH-316 SS	280	350
1.4	Inconel718	400	220	2.4	PH-316 SS	400	220
1.5	Inconel718	400	300	2.5	PH-316 SS	400	300
1.6	Inconel718	400	350	2.6	PH-316 SS	400	350
1.7	Inconel718	750	220	2.7	PH-316 SS	750	220
1.8	Inconel718	750	300	2.8	PH-316 SS	750	300
1.9	Inconel718	750	350	2.9	PH-316 SS	750	350

Table 1.

DOE case setup for CFD simulations.

In addition to the defined input variables for the case study, i.e., material type, laser power and speed, many other parameters are also needed to setup the CFD simulation in FLOW-3D AM®. The simulation setup is quite critical as the melting process itself is very complex. Basic setup steps are described here. Interested readers can visit here to obtain more in-depth knowledge and tutorials to use for individual studies.

• In FLOW-3D AM® start a melting simulation in workspace, using CGS unit system and kelvin for temperature.

• In global settings, set finish time for simulation. We have used 0.0005 s for the initial simulations. But it can be varied as per desired setup.

• Bubble and phase change physics activated with constant pressure of vaporization where

  • Pressure = 1.0e6 dyne/cm²,

  • Heat transfer co-efficient =1.0e5 erg/cm²/s/K

  • Gamma = 1.4.

• Density is set as function of temperature.

• Gravity, in Z direction, −981 cm/s².

• Surface tension, laminar viscous flow and solidification mode activated.

• Surrounding fluid is air at 15°C.

• Meshing size of cell 0.0005 cm.

• The simulation area can be created using particle size, density, maximum number of particles, packing density etc. using FLOW 3D and then converted into STL into particle to STL converter. This STL file can be used as base for all case studies as “particle bed”. Additional fluid region can also be incorporated.

• Laser power, velocity, lens shape and size can be defined using FLOW 3D WELD module. In this study we have used circular lens with focal distance 0.8 cm, radius of 0.01 mm and spot radius of 0.003 cm. Only X directional velocities are used, they are zero in Y and Z directions.

• Materials are used only Inconel-718 and PH-316 Stainless Steel in this case study, and the built-in property directory is utilized as well. In specific studies, we can also edit any relevant properties of material as needed.

6. Results and discussion

The results obtained from FLOW 3D AM can be analyzed using simple rendering in the program itself or can be further post processed using their FLOW 3D POST package. It is necessary to include/activate the required outputs whilst setting up the simulation, so that it is easier to post process. After completion of simulations, usually the flsgrf.melting files are rendered and used in further post processing. As we have several cases in the DOE, we are only discussing a few sample cases in this paper due to limitation of space.

In Figure 10, we are attempting to provide a visualization of how the laser moves starting at the left side and the temperature distribution changes with time. We are including only 6 sample timesteps to demonstrate the progress. In Figure 10(a), we capture the time step at t = 0.000005 s. The progression occurs through Figure 10(d), at t = 0.000305 s, where the laser has reached the end with melt pool at highest temperature. Moving further in time, on Figure 10(e) at t = 0.0004 s, we can see the melt pool gradually is cooling off. This is clearly visible at the right.

Figure 10.

Laser progression and temperature profile map (a) to (f) shows progression in time domain.

Next, in Figure 11, the 2D cross section (X-Z plane) is presented on progressive time step, from top to bottom. Only few timestep captures are only provided here for ease of conception. Important feature to notice here is in such a 2D view, we can easily identify the voids or porous structures that are left behind during the melting process. With this knowledge we can further investigate to correlate these to defect generation and thus optimize process parameters accordingly. In this technique, studying the microscopic density, energy density etc. will also contribute profoundly and these spatio-temporal data are also available.

Figure 11.

Representative melt pool on 2D plane (progression from top to bottom).

7. Future work

The work described in this paper is a part of research project that aims at developing an intelligent closed loop feedback-controlled 3D metal printing system. The simulation framework will be used to generate training data for neural net training in addition to real life experimental results. With sufficient training data, the developed neural network can serve as a significant tool to minimize anomalies in metal additive manufacturing. These will be the foundation of the automated printing system that will choose the process parameters in order to minimize defects and deformities. This project will be a new approach in metal printing industry that will ultimately improve the quality and quantity of product.

8. Conclusion

Additive manufacturing has been in practice for decades, yet continuous developments are being carried out due to its complex nature. In this paper, a modeling framework is presented to help understand the physical phenomenon behind the printing technique and correlate the process parameters with generated defects. The framework includes the computer numerical modeling techniques to be used to simulate the macro and micro scale phenomenon of powder bed fusion process. This is designed in a way to identify temperature signatures and anomaly development occurrences to help identify the root causes of defect generation and in a way categorize the optimal process parameters. The next step includes experimental validation of the techniques. This paper contains relevant information on the experimental testing on industrial printing bed. Being in the initial stage of the project, though we have not completed the experimental studies, and we believe any miniature scale experimentation would not do justice to the presented framework. We expect the context provided in this paper will help research and industry community to improve the state of the art in this field. Industry needs minimum defects and maximum quality with sufficient reliability to be able to flourish in international market. The authors hope this paper would help metal additive manufacturing industry in this matter.