AM PravaH: 3D Printing Software By Paanduv Applications

-Dr. Rimzhim Gupta (Research Scientist at Paanduv Applications)

About the case study

This document will help you run your first AM PravaH LPBF simulation for macroscale and microstructure modeling. The case study is done for a standard alloy of titanium i.e. Ti6Al4V widely used for aerospace and biomedical applications. The simulation is carried out for a multilayer, multi-track scan pattern. The process parameters for the laser are e.g. 300 W power and e.g. 1 m/s scan speed. D4 sigma or spot dia is 0.1 mm. 

Exclusive features of AM PravaH include consideration of 4 phases, with no explicit formulation for recoil pressure because evaporation and recoil pressure effects are included in the vapor phase. AM PravaH takes the thermophysical properties of the 4 phases as inputs along with the process parameters such as laser power, laser scan speed, shielding gas flow angle and velocity, layer thickness, spot diameter, preheating temperature, chamber initial pressure, and substrate initial height, scan pattern, particle size distribution and number of layers. Macroscale modeling will generate the following outputs melt pool dimensions, porosity %, thermal gradients, cooling rates, and thermal cycling plots. The Microstructure module generates outputs such as grain size distribution, angular chord length distribution, misorientation angles, and Euler angles. AM PravaH leverages the amalgamation of computational modeling solvers performing multiphase macroscale modeling, microstructure modeling, and Artificial Intelligence (AI) at a unified platform.  

Keywords: Additive Manufacturing, 3D printing, Meltpool dynamics, porosity, thermal gradients, cooling rate, microstructure analysis, grain size, angular chord length, 3D printing software 

AM PravaH: 3D printing simulation software 

Introduction

AM PravaH is a 3D printing simulation software that facilitates end-to-end solutions for Additive Manufacturing physics-based simulations. Additive Manufacturing is commonly known as 3D printing. This 3D printing software is the “World’s first all-inclusive 3D computational software for Additive Manufacturing”. This 3D printing software will be useful for new alloy development, process parameter optimization, and reducing defects. Not to confuse with other 3D printing software that only performs thermomechanical analysis and deformation, AM PravaH software offers much more.

The software capabilities in brief are as follows:

There are three modules in AM PravaH

(i) Macroscale modeling

(ii) Microstructure analysis

(iii) Integrated Deep Learning (AI) module 

Relevance of the 3D printing software; AM PravaH

This 3D printing software is relevant for researchers, academicians, and industry professionals from large and small-sized companies working in the Additive Manufacturing field. Where the primary focus is to understand the in-depth physics, microstructures, and effect of process parameters on the melt pool dynamics and distribution and the root cause of the defects. If we have a much closer look at the fundamental processes of 3D printing; this is a transient, very quick, and very dynamic process, which can't be captured with the naked eye. Therefore, AM PravaH simulations can be extremely useful in understanding this part. 

Based on that, one makes important decisions such as 

1. Which alloy is better?

2. How do we get solidification, melting, and cooling rates information?

3. what are the best operating process conditions to get a defect-free and dense part

4. What do the microstructures look like?

5. What are the grain distribution and mean grain size? 

6. What are the melt pool sizes?

7. How much will be the porosity %?

Important notes

1. This software is not relevant for folks seeking help in support design and optimization, topology optimization, slicing, and CAD work assistance. 

2. Although this document does not show the use of the AI solver of AM PravaH is a non-linear regressive model that is flexible with the data, it is robust. It can be a powerful way to utilize your data and predict important details such as materials properties, process parameters, mechanical properties, residual stresses, output parameters such as porosity, density, cooling rates, thermal gradients, melt pool width, depth, height, and more. 

AM PravaH home page

Once you open the AM PravaH home page, you have to save your project by the name “dummy”. Once the project is saved at the desired location, you can start with the module that you would like to choose (i) macroscale, (ii) microstructure, and, (iii) AI.

When you run your first case in the macroscale module, the prerequisites are (i) knowing the thermophysical properties of your material and (ii) the process parameters you want to use. 

Once you know these inputs, you are all set to run your first AM PravaH simulation.

To achieve the most accurate results using any 3D printing simulation software, the first step is calibration. Additive Manufacturing is a complex possess that uses multiple parameters at the same time and all the equations are not analytically accurate. Some empirical correlations may need coefficients that may vary in a large range. These parameters may need tweaking before running a full-fledged simulation. These parameters in the case of AM PravaH are Fresnel coefficients and Liquid absorptivity. Hence, calibration i.e. comparing a single-layer melt track using the literature or experiments is a wise idea to ensure the accuracy of other high-fidelity simulations.

The flow chart (Figure 1) depicts the step-by-step procedure to run a successful simulation using AM PravaH. 

Figure 1. Schematic representing the step-by-step procedure to run the first AM PravaH case study

Steps to run the first AM PravaH case study 

1.Calibration

You start with any software or an instrument, the first thing that you do is calibrate. It ensures that the further measurements that you will perform will be accurate. Run the first simulation by entering the default settings in AM PravaH, and compare your results with the experimental data (this may come from any research article or directly from a 3D printing machine). The schematic below Figure 2 briefs about this procedure in the form of a flow chart. 

Figure 2. Snippet from the AM PravaH software for Fresnel Coefficients settings, Detailed flow chart including the calibration and parameter optimization 

If the errors are there adjust parameters such as liquid-vapor & liquid-gas Fresnel Coefficients in process settings. Change these parameters in the same way (0-1) to get the accurate melt pool parameters, as shown in Figure 3. Once the results agree with the experiments, the solver is calibrated and you can proceed for further analysis.

Figure 3. Snippet from the AM PravaH software for Fresnel Coefficients settings

We have done the calibration for Ti6Al4V using a research article [1] where we compared the default settings of AM PravaH first. We changed the Fresnel Coefficients values of liquid-vapor and liquid-gas to 0.15, 0.15 from 0.08 and 0.08 and we approached Melt pool depth and width errors 1.76 and 6.16 %, as shown in the table below (Table 1).

All the further simulations are carried out at Fresnel coefficients 0.15 for liquid interfaces.

2. Setting up the macroscale module case 

Macroscale modeling of AM PravaH is much needed for new alloy development, building a quality defect-free and dense part, and for process parameter optimization. Using coupled multiphysics such as 4-phase multiphase modeling (solid, molten metal, metal vapor, and inert gas), heat and mass transfer, laser dynamics, particle dynamics, and Marangoni convection are rightfully used in the macroscale model. The use of 4 phases in the model leaves no room for assumptions and hence ensures accuracy. The model takes inputs in three forms (i) thermophysical properties (ii) process parameters and (ii) simulation settings. 

Figure 4. 3D printing software; AM PravaH snippets for the input settings of the macroscale module

More details about the individual settings are covered in the sections below.

2.1 Material properties

It's important for users to know the thermophysical properties to run AM PravaH. But here is the great news! we have provided you with the database of 14 material properties which are Ti6Al4V, IN718, IN625, IN738L, SS316, AlSi12, Hastelloy-X, Maraging steel, Al5083, Cu-ETP, CuCrZr, SS316L, Ti6242, and SS304. Apart from this, the best thing is these values are editable, so if you have developed your own alloy from one of these you can change these properties. That’s not it, you can in fact enter a new material that is completely unknown or not out of these alloys and save those properties. That will be saved in your local machine as a new alloy in the GUI for the next run. Isn’t it amazing?

After selecting the material or alloy you wish to model, the next thing is selecting the shielding gas out of argon, nitrogen, and air. 

Once you select the material and shielding gas, we have the next interesting selection for you. We are giving users the flexibility to model 3-phase simulations or 4-phase simulations. This is done by choosing ON or OFF in the drop-down menu in the vaporization option. If the user knows that the laser power is low (< 150 W), the alloy may not reach the vaporization temperature hence no vapor will be formed then why solve for vapor? The three-phase simulation will be much faster. But it is always recommended to do the first run with vapor ON i.e. 4-phase simulations, that way you can estimate how far the temperatures are going 

Figure 5. AM PravaH snippets for entering the thermophysical properties as materials input settings 

For this simulation, we used the parameters from the calibration step and used Ti6Al4V alloy Argon gas as shielding gas with the same properties from the AM PravaH database and kept the vaporization ON.

2.2 Process parameters

The second important input that the users should know is the process parameters. This includes the machine & laser parameters direction of the laser, laser power, D4 sigma or spot diameter, Fresnel coefficients, layer thickness or spreader clearance, spreader velocity,  initial chamber conditions (preheating T, initial chamber pressure, initial substrate height), shielding gas velocity and gas flow angle, scan pattern, number of layers to model and the particle size distribution. Believe me, Additive Manufacturing processes take these many parameters into account.

Figure 5. AM PravaH snippets for entering the thermophysical properties as materials input settings

More details about the scan pattern and particle size distribution, you will know in the following sections. 

Other process parameters used in the simulation are 266 W laser power, 1 m/s scan speed, 0.1 mm spot diameter, 60 micro-meter layer thickness 50 degrees C preheating temperature, Where there was no shielding gas flow.

2.2.1 Scan pattern

Giving the scan pattern in AM PravaH requires some knowledge of the coordinate system. The user needs to upload the san pattern file in the form of .csv format. The .csv will be saved in the datasets folder of your project folder. It has 6 columns 1st, number of layers, 2nd, time, 3rd, 4th, 5th, x, y, z coordinates of the laser projection on the substrate or powder bed, and 6th laser ON/OFF condition. 

Details of the scan pattern for this simulation include 3 number of layers, a hatch distance of 120 microns, a hatch offset of 200 microns, and a hatch delay time of 0.6 ms. This strategy is depicted in Figure 6.

Figure 6. (i) Schematic depicting the scan pattern writing strategy in AM PravaH, (ii) tabulated .csv file, and (iii) 3D scan pattern visualization using AM PravaH 

Here, the 2nd column time is still left to understand. This is the time at which the laser will be at those corresponding (X, Y, Z) points. 

Most importantly, the scan speed is neither explicitly mentioned nor asked by AM PravaH. It is calculated using the scan pattern. How? Let's understand this.

When you mention the x-coordinate (start X0 and endpoint X1) and the time at which you want to take the laser (from time =0 to t), the scan speed is (X1-X0)/(t-0). Figures 6 (i), (ii), and (iii) indicate how the coordinate system of the scan track can be converted into a .csv format that is an input file for AM PravaH. Further, you can also visualize and cross-check if the scan pattern file is accurate using the “visualize scan pattern” tab in the GUI as given below.      

Figure 6. (i) Schematic depicting the scan pattern writing strategy in AM PravaH, (ii) tabulated .csv file, and (iii) 3D scan pattern visualization using AM PravaH 

Here, the 2nd column time is still left to understand. This is the time at which the laser will be at those corresponding (X, Y, Z) points. 

Most importantly, the scan speed is neither explicitly mentioned nor asked by AM PravaH. It is calculated using the scan pattern. How? Let's understand this.

When you mention the x-coordinate (start X0 and endpoint X1) and the time at which you want to take the laser (from time =0 to t), the scan speed is (X1-X0)/(t-0). Figures 6 (i), (ii), and (iii) indicate how the coordinate system of the scan track can be converted into a .csv format that is an input file for AM PravaH. Further, you can also visualize and cross-check if the scan pattern file is accurate using the “visualize scan pattern” tab in the GUI as given below. 

Figure 7. Scan track visualization and check scan speed option in AM PravaH

2.2.2 Particle size distribution

Usually, particle size distributions are described using d10, d50, and d90. This means that 10 %, 50 %, and 90 % of particles are below the corresponding numbers in diameter. In this case, d10 27-32 d50 44-48 d90 70-75, this means that 10 % of the particles are of diameter between 27-32 𝞵m, 50 % of the particles are of diameter between 44-48 𝞵m and 90 % of the particles are of diameter between 70-75 𝞵m. However, Dont’t be confused! AM PravaH takes the particle size distribution input in radius (m).  The table below (Figure 8) shows different fractions of the particles ranging from 27-75 𝞵m.

Figure 8. Input .csv file for particle size distribution in AM PravaH

2.3 Simulation parameters

The simulation details for simulation settings are shown below in Figure 9 as follows: Mesh Size=4.5e-06, Courant Number=0.5, Max time step=2.25e-06

In this case, the higher the ACCURACY, the smaller the mesh size, and hence, the more accurate will be the simulation at the expense of significantly higher simulation time. The smaller the performance, Conversely, adjusting the PERFORMANCE slider increases the mesh size, optimizing computational efficiency.

Figure 9. Simulation settings in AM PravaH for macroscale modeling

3. Running the case

The case setup in AM PravaH is pretty straightforward, just after saving all the settings. Hit RUN SOLVER. You can also check the powder spreading process and can check the status of the macroscale modeling using LAUNCH CONSOLE, while will open the residuals. Now you can also resume your simulation after saving the simulations. In case your simulation stops for whatever reason, you can resume it. Don’t worry, your data is not lost.

Figure 10. Execution options in AM PravaH GUI

4. Postprocessing of macroscale modeling simulation

AM PravaH provides a state-of-the-art postprocessor “Paraview” linked with the GUI which will help you observe and analyze the results even when the simulation is running and not completed. The interesting part of this DEM modeling is that this also applies to multilayer LPBF simulation. In a three-layer simulation, after the solidification of the first and second layers, it takes the solidified tracks as the substrate and starts spreading the particles accordingly. Hence it applies to the layers even more than 3 number of layers. You can go up to the number of layers your computers allow.

1. You can visualize particle spreading

2. You can visualize the complete solid in a 3D view

3. You can visualize the melting, solidification, and vaporization separately or simultaneously and take animation and screenshot

4. You can quantify the individual fractions of 4 phases (solid, liquid, vapor, and shielding gas), density, temperature, pressures, and velocities. You can plot that directly in Paraview or export the raw data. 

5. You take slices, and 3D clips of the complete domain to visualize the porosity, defects, and melt pools in the best possible way. 

6.  You can plot thermal cycling plots, temperature gradients, and cooling rates at a single location or multiple locations (using probes) and evaluate porosity using this as shown in the figures below.

Figure 11. Post-processed data from the AM PravaH simulation (a) Powder spreading and 2 solidified layers in LPBF simulation (initiation of the third layer) (b) Powder spreading of the third layer on top of 2 solidified layer (c) thermal cycling plot (d) porosity evaluation (e) thermal gradients (f) cooling rates.

5. Setting up the microstructure module case 

Microstructure modeling is done using a stochastic approach of the Monte Carlo Potts algorithm. Monte Carlo Potts model is an on-lattice technique for the simulation of curvature-driven grain growth. While the heat source is not directly simulated, its effect is imposed as a molten zone surrounded by a high-temperature, heat-affected zone (HAZ) having a steep thermal gradient. Together, the molten and heat-affected zones provide the kinetics necessary to enable microstructural evolution.

The Microstructure module of AM PravaH does not require too many thermophysical parameters into account as it only performs a conduction-based heat transfer process, hence considering the entire domain as solid, NO LIQUID. The required inputs in AM PravaH are properties such as solidus temperature, liquidus temperature, and thermal diffusivity. In process parameters, the software requires scan speed, laser power, scan pattern, melt pool adjustment parameters, laser absorption fraction, etc. 

5.1 Adjustment parameters 

The rest of the parameters are fine, one may think what are these adjustment parameters? Do we need to adjust the melt pool? Well, the user will get the melt pool dimensions either from experiments or from macroscale modeling simulations of AM PravaH. According to the equation given below the microstructure model creates a semi-circular melt pool cross-section, if ny and nz are equal, which is not usually the case. Mostly the melt pool cross sections are elliptical as shown in figure 12.

Equation

Figure 13. Analysis of the melt pool width and depth in Paraview (No. of lattice site * length of one lattice site; 65*1.66 micro-meter)

6. Running the case

Running a microstructure case is very simple. Once you have entered the material properties and process parameters. Just adjust the computational settings (accuracy, performance, data saving frequency), and spin states (Initial number of random states).  You might as well reduce the data-saving frequency to the lowest because it may eat up your good enough space in your computer. Just hit run solver now!

7. Postprocessing of microstructure modeling simulation

You can check your results during or after the simulation is completed. You can check the estimated simulation time and space required in the terminal by hitting “Launch Console”. When you do this, the data will be converted into the visualization mode. It may take a few minutes.

With this data, you can visualize temperature distribution, Euler angles, grain evolution, grain misorientation, etc. 

Paanduv has developed its own data analysis and visualization that helps metallurgists and material scientists gather more meaningful information such as the percentage of columnar and equiaxed grains and respective grain size distribution, overall mean grain size distribution, angular chord length distribution, or polar plots. That looks something like Figure 15.

Figure 14. The window of GUI further data analyses can be done (Ignore consecutive cells = 3), which means that it will consider at least 3 cells of the same spin as one grain  

     Figure 15. (a) Grain evolution in x, y, and z planes, (b) polar plots in x, y, and z planes and (c) mean grain size distribution 

8. Correlating the post-processed data with mechanical properties

AM PravaH evaluates the % of columnar and equiaxed grains and also gives the option to use the Hall-Petch equation to predict parameters such as Fatigue, Hardness, Yield strength, and Crack propagation. 

9. What's new and what’s in the pipeline?

(vi) Faster AM PravaH → Basis preemptive time steps and numerical schemes for different physics-based phenomena the timescale could be reduced by more than half. [2]

(vii) More accurate AM PravaH → Inclusion of temperature-dependent thermophysical properties will take you closer to the more accurate solution. Also with the help of advanced numerical and meshing techniques, the solvers are more stable and hence approaching better accuracy. 

References

[1] Ransenigo, Chiara, et al. "Evolution of melt pool and porosity during laser powder bed fusion of Ti6Al4V alloy: numerical modeling and experimental validation." Lasers in Manufacturing and Materials Processing 9.4 (2022): 481-502.

[2]https://www.linkedin.com/pulse/preventive-approach-healthy-numerical-simulations-adwaith-gupta-o71bc/?trackingId=nulAYXSpTMCUKJbrl52E3g%3D%3D