Accelerating Part-Scale Simulation in Liquid Metal Jet Additive Manufacturing via Operator Learning

Søren Taverniers,Svyatoslav Korneev,Kyle M. Pietrzyk,Morad Behandish

DOI: https://doi.org/10.48550/arXiv.2202.03665

2022-02-03

Abstract:Predicting part quality for additive manufacturing (AM) processes requires high-fidelity numerical simulation of partial differential equations (PDEs) governing process multiphysics on a scale of minimum manufacturable features. This makes part-scale predictions computationally demanding, especially when they require many small-scale simulations. We consider drop-on-demand liquid metal jetting (LMJ) as an illustrative example of such computational complexity. A model describing droplet coalescence for LMJ may include coupled incompressible fluid flow, heat transfer, and phase change equations. Numerically solving these equations becomes prohibitively expensive when simulating the build process for a full part consisting of thousands to millions of droplets. Reduced-order models (ROMs) based on neural networks (NN) or k-nearest neighbor (kNN) algorithms have been built to replace the original physics-based solver and are computationally tractable for part-level simulations. However, their quick inference capabilities often come at the expense of accuracy, robustness, and generalizability. We apply an operator learning (OL) approach to learn a mapping between initial and final states of the droplet coalescence process for enabling rapid and accurate part-scale build simulation. Preliminary results suggest that OL requires order-of-magnitude fewer data points than a kNN approach and is generalizable beyond the training set while achieving similar prediction error.

Fluid Dynamics,Artificial Intelligence,Machine Learning,Computational Physics,Data Analysis, Statistics and Probability

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to accelerate part - scale simulations through the Operator Learning (OL) method in the Liquid Metal Jet Additive Manufacturing (LMJ) process. Specifically, when traditional high - fidelity numerical simulations are used to predict part quality in the additive manufacturing process, partial differential equations (PDEs) describing multi - physical fields need to be solved, which is very computationally intensive at the part scale, especially when many small - scale simulations are required. In the paper, liquid metal jetting (LMJ) is taken as an example to illustrate this computational complexity.

To reduce computational costs, previous studies have constructed some reduced - order models (ROMs) based on neural networks (NN) or k - nearest neighbor (kNN) algorithms. These models can replace the original physics - based solvers and are computationally feasible in part - level simulations. However, although these models are fast in inference, they often sacrifice accuracy, robustness, and generalization ability.

Therefore, this paper proposes a new method, that is, applying operator learning (OL) to learn the mapping between the initial state and the final state of the droplet coalescence process, thereby achieving fast and accurate part - scale construction simulations. Preliminary results show that OL requires an order of magnitude fewer training data points than the kNN method, and can generalize beyond the training set while achieving similar prediction errors.

The core contribution of the paper lies in using operator learning techniques, especially the Fourier Neural Operator (FNO), to improve the speed and accuracy of part - scale simulations in the liquid metal jet additive manufacturing process while reducing the need for a large amount of training data. This method not only improves the efficiency of simulations but also enhances the generalization ability of the model, enabling it to better handle parts of different shapes and structures.