Bayesian updating for thermal model of laser powder-bed fusion additive manufacturing

This study aims to enhance the accuracy of thermal models for laser powder bed fusion in metal additive manufacturing (AM) by addressing uncertainties related to stochastic model parameters and inherent biases. Given the limited availability of experimental data due to resource constraints, a rigorous quantification of cognitive uncertainties is essential. The proposed methodology employs Bayesian model updating theory to calibrate an existing thermal model, incorporating an efficient algorithm known as Bayesian Updating with Structural Reliability Method (BUS) to execute the calibration task. Additionally, a Gaussian process regression surrogate is integrated to further mitigate the computational burden. This comprehensive approach facilitates robust uncertainty quantification, particularly concerning the thermal pool w idth and depth. Despite the complexity of physical mechanisms and high computational costs, this research seeks to improve the predictive capabilities of simulation models, thereby advancing the understanding of the "process-structure-property" relationship in selective laser melting technology.