Integrating Finite Element Based Heat Transfer Analysis with Multivariate Optimization for Efficient Weld Pool Modeling

Abstract

Measurement of weld thermal cycle and peak temperature is a difficult task in fusion welding due to high peak temperature and rapid thermal cycle. Numerical modeling of heat transfer process in fusion welding provides a useful tool for the prediction of the weld pool shape and thermal cycles. However, the reliability of the predictions greatly depends on the accuracy of the input parameters provided to such models. In the case of laser beam welding, the absorptivity represents an important variable that defines the actual heat input for a given power while the value of absorptivity rarely known a-priori with confidence. The present work provides a novel framework where the unknown value of the absorptivity is obtained using an inverse approach by integrating finite element based heat transfer simulation and multivariate optimization procedure. The heat transfer simulation predicts the weld pool dimensions for known welding conditions and assigned values of absorptivity. The optimization algorithm tracks the error in the prediction, its sensitivity with small change in absorptivity, and finally yields the optimum value of absorptivity iteratively. Eight known welding conditions and corresponding weld pool measurements were utilized in this work. It is experienced that the optimum value of absorptivity remains nearly same irrespective of the number of known measurements used for the optimization process thereby indicating the robustness of the inverse approach.