Understanding where and how simulation applies to metal 3D printing, including its limitations.
Metal additive manufacturing (AM) generally involves highly complex parts made from high-cost materials, with more than a hundred parameters affecting process and fabrication quality. For these reasons, a trial-and-error process of test prints is often time-consuming and costly. Fortunately, engineering simulation can help ensure the success of metal 3D print jobs faster and at a lower cost.
What types of metal AM can be simulated?
Metal AM processes can be classified according to their energy source and material form into the following three categories:
- Binder Jetting, which uses a liquid binding agent on metal powders
- Powder Bed Fusion, which uses lasers or electron beams on metal powders
- Directed Energy Deposition, which uses lasers or electron beams on metal powders or wire
Of these three categories, the majority of simulation software developed has focused on powder bed fusion (PBF) and directed energy deposition (DED). This is not to suggest that there are no simulation software options for the other processes – for example, Hexagon’s Simufact Additive and Desktop Metal’s Live Sinter both have capabilities for simulating metal binder jetting – but the breadth of options and depth of research into PBF and DED simulation is considerably more advanced.
How does simulation help metal AM?
As indicated above, simulation reduces both the time and the cost of 3D printing metal parts. More specifically, simulation software can predict hot spots and lack of fusion, both of which reduce the material strength of 3D printed parts. Simulation can also be used to improve the metal 3D printing process itself by (in the case of powder bed fusion) identifying and correcting recoater interference, which occurs when parts exceed layer height or do not adhere to the substrate. Finally, and perhaps most importantly, simulation can predict support structure failures and help engineers understand where and how to adjust the placement of support structures to ensure successful prints.
What are the focus areas for metal AM simulations?
Because metal additive manufacturing involves complex physical phenomena occurring across multiple time- and length-scales, simulations for metal AM tend to be somewhat narrower than for other manufacturing processes. Typically, a simulation of a metal AM process will focus on one or possibly two of the following, at most:
- Microstructure evolution
- Melt pool behavior
- Residual stress
- Crack propagation
In the last few years, it’s become common practice to classify metal AM simulation models for these focus areas according to the scale at which they operate: macro, meso and micro. The finite element method (FEM) is the standard approach for analyzing residual stresses and mechanical deformations at the macroscale, though the finite volume method (FVM) has begun to see more use in recent years. At the mesoscale, computational fluid dynamics (CFD), smoothed particle hydrodynamics (SPH) and the lattice Boltzmann method (LBM) are used to analyze fluid flow in the melt pool. And, at the microscale, cellular automata (CA), Monte Carlo and phase field (PF) models are most useful for analyzing grains, nucleation and voids.
What are the limits of simulation for metal AM?
Due to the large number of parameters and complex multi-scale physics involved in metal additive manufacturing, simulating an entire part over the entire printing process is generally not computationally feasible.
Early metal additive manufacturing simulations involved thermal models based on FEM. These were similar to the thermal models in computational welding mechanics, where laser/material interactions are simulated by a moving heat flux or volumetric heat source with a planar Gaussian profile.
These are useful as a first approximation, especially since they’re computationally inexpensive, but they also involve coarse assumptions about the metal AM process because they don’t account for fluid dynamics. As a result, these tend to overestimate melt pool temperature because they neglect the convective heat transfer from fluid dynamics.
In more recent years, the closest engineering and materials science researchers have come to total AM part simulations is coupling CFD simulations of the melt pool to subsequent solid mechanical models that take thermal input from the CFD model. However, this is still computationally intensive and so has only been carried out at the mesoscale level with one or two tracks.
