Prof. Heping CHEN

Texas State University, USA

Topic

Optimal Welding Parameter Learning for Robotic Arc Welding Process

Abstract

    Robotic arc welding has been implemented in many applications for manufacturing automation. Once a robotic welding system is set up, one of the challenges is to tune the welding process parameters. Some methods have been proposed to explore optimal welding parameters, however they are suffering some practical limitations. Because the welding process parameters determine the weld quality, it is important to find optimal set of parameters to achieve best weld quality. In this talk, a welding process parameter optimization method is presented to optimize weld quality for robotic arcwelding process. A non-parametric model is constructed using Gaussian Process Regression iteratively while Bayesian Optimization Algorithm is adopted to exploit potential optimal candidate. To evaluate the weld quality, a tensile machine is utilized to measure the tensile strength of the weld joint. Experimental results show that an optimal set of welding process parameters can be determined using ten trials. The number of experiments to explore optimal welding process parameter is greatly decreased compared with the existing methods. Hence the proposed method is an efficient and effective tool to optimize weld quality for complex welding process.