A Physics-Informed Two-Level Machine-Learning Model for Predicting Melt-Pool Size in Laser Powder Bed Fusion

Abstract

Laser powder bed fusion (L-PBF) additive manufacturing (AM) is one type of metal-based AM process that is capable of producing high-value complex components with a fine geometric resolution. As melt-pool characteristics such as melt-pool size and dimensions are highly correlated with porosity and defects in the fabricated parts, it is crucial to predict how process parameters would affect the melt-pool size and dimensions during the build process to ensure the build quality. This paper presents a two-level machine learning (ML) model to predict the melt-pool size during the scanning of a multi-track build. To account for the effect of thermal history on melt-pool size, a so-called (pre-scan) initial temperature is predicted at the lower-level of the modeling architecture, and then used as a physics-informed input feature at the upper-level for the prediction of melt-pool size. Simulated data sets generated from the Autodesk's Netfabb Simulation are used for model training and validation. Through numerical simulations, the proposed two-level ML model has demonstrated a high prediction performance and its prediction accuracy improves significantly compared to a naive one-level ML without using the initial temperature as an input feature.

Progress and Challenges of Additive Manufacturing of Tungsten and Alloys as Plasma-Facing Materials

“…Previous methods for simulating this behavior include finite element analysis (FEA) modeling and fluid mechanics; however, they are problematic due to high-power infrastructure requirements for proper utilization [160]. ML was used to properly predict melt pool size as a function of the AM process parameters [159]. Similar results were obtained when employing a neural network model to predict the temperature and melt pool fluid dynamics in metal AM processes [161].…”

“…Simulations of melt pool behavior are of interest in AM due to their direct correlation with the behaviors of fabricated components [158]. Ren et al employed a two-stage ML model to predict melt pool behavior during AM scanning [159]. Previous methods for simulating this behavior include finite element analysis (FEA) modeling and fluid mechanics; however, they are problematic due to high-power infrastructure requirements for proper utilization [160].…”

Multiphysics multi-scale computational framework for linking process–structure–property relationships in metal additive manufacturing: a critical review

“…In another attempt, a PIML model based on Neural networks coupled thermal sensing and FEM simulations for porosity prediction in a LDED process [498]. Further, a Multilayer perceptron model incorporating FEM transient thermal fields predicted correlation between scanning patterns and thermal history for a LDED process [499]. In the LPBF process, a powder spreading process map was created using a DEM-based powder spreading model coupled with a back-propagation Neural Networks [500], the corresponding PIML model framework is illustrated in Figure 46.…”

Data-Driven Prediction and Uncertainty Quantification of Process Parameters for Directed Energy Deposition

“…However, deterministic models cannot provide uncertainty quantification (UQ), which is crucial for reliable additive manufacturing due to the various sources of uncertainties in additive manufacturing [30][31][32][33]. Probabilistic machine learning models such as Gaussian Process Regression (GPR) [34,35] can account for this UQ and have been applied in laser powder bed fusion (LPBF) to predict the melt pool geometry [36][37][38][39][40][41] or in DED to predict the mechanical properties [42], the component height [43], the geometry of single tracks [44,45], or melt pool geometry [46,47] based on the process parameters. The inverse problem, i.e., the determination of a suitable process to produce the desired track geometry, can principally be solved by combining the regression model with an optimization algorithm.…”