Thermal field prediction for laser scanning paths in laser aided additive manufacturing by physics-based machine learning

Abstract

Laser aided additive manufacturing (LAAM) is a key metal 3D printing and remanufacturing technology for fabrication of near-net shape parts. Studying thermal field induced by different scanning strategies is important to evaluate and optimize the resultant residual stress and distortion distribution. However, it is very computationally expensive to simulate multi-bead deposition process using existing numerical model to analyze and select appropriate laser scanning strategies. In this paper, we make use of a recently developed and experimentally validated efficient thermal field prediction numerical model for LAAM to generate training data for a physics-based machine learning algorithm. A combined Recurrent Neural Networks and Deep Neural Networks (RNN–DNN) model was developed to identify the correlation between laser scanning patterns and their corresponding thermal history distributions. Subsequently, the developed RNN–DNN model is able to make thermal field prediction for an arbitrary geometry with different scanning strategies. Comparison between the numerical simulation results and the RNN–DNN predictions showed good agreement of more than 95%.