Novel Calibration Strategy for Validation of Finite Element Thermal Analysis of Selective Laser Melting Process Using Bayesian Optimization

Abstract

Selective laser melting (SLM) produces a near-net-shaped product by scanning a concentrated high-power laser beam over a thin layer of metal powder to melt and solidify it. During the SLM process, the material temperature cyclically and sharply rises and falls. Thermal analyses using the finite element method help to understand such a complex thermal history to affect the microstructure, material properties, and performance. This paper proposes a novel calibration strategy for the heat source model to validate the thermal analysis. First, in-situ temperature measurement by high-speed thermography was conducted for the absorptivity calibration. Then, the accurate simulation error was defined by processing the cross-sectional bead shape images by the experimental observations and simulations. In order to minimize the error, the optimal shape parameters of the heat source model were efficiently found by using Bayesian optimization. Bayesian optimization allowed us to find the optimal parameters with an error of less than 4% within 50 iterations of the thermal simulations. It demonstrated that our novel calibration strategy with Bayesian optimization can be effective to improve the accuracy of predicting the temperature field during the SLM process and to save the computational costs for the heat source model optimization.