Robotic welding parameter optimization based on weld quality evaluation

Abstract

Arc welding is widely used in industrial applications. In arc welding process, welding parameters significantly affect weld quality such as mechanical properties. However it is very challenging to tune these parameters to achieve best weld quality. Some researchers developed some methods to optimize welding parameters, but these methods have many limitations. In this paper, we propose an innovative method to optimize welding parameters for robotic welding process using Gaussian Process Regression (GPR) and Bayesian Optimization Algorithm (BOA). The relationship between the weld quality and welding parameters is modeled based on GPR. To expedite the optimization process, BOA is applied to balance the modeling process and optimization process. The Gas tungsten arc welding (GTAW) process is utilized to test the proposed method. The tensile strength of the welded joints was measured to evaluate the weld quality. The experimental results demonstrate that the proposed method can find a set of optimal welding parameters using about ten experiments. Thus it greatly improves the welding parameter optimization process by reducing the number of trial experiments compared to the existing methods such as Design-of-Experiments. The proposed method can also achieve better solution by evaluating more parameter values and exploring the whole parameter space. Therefore it provides an efficient and effective tool to optimize welding process.