Modeling of human welder response to 3d weld pool surface: Part II -Results and analysis

Understanding and modeling a human welder's responses to a 3D weld pool surface may help develop intelligent welding robotic systems and train welders faster. In this first effort on modeling a human welder's behavior, a novice welder's adjustment on the welding current as a response to the 3D weld pool surface as characterized by its width, length, and convexity is studied. The first part of the paper used an innovative machine vision system to measure/record in real time the specular 3D weld pool surface from experiments and conducted preparation experiments to reduce the inconsistencies in the welder's responses as well as determine the welder's delay and time intervals of the process response. In this part of the paper, experiments are designed to produce random changes in the weld pool using random welding speeds in order to model the response of the welder to a dynamic weld pool surface. The fluctuating weld pool surface and welder's adjustments on the welding current are recorded. Through the least squares algorithm, various models with different structures are identified to correlate the current adjustment to the 3D weld pool surface. It is found that the human welder's responses are not only related to the 3D weld pool surface but also rely on the welder's previous adjustments. The resultant model has been verified by further experiments for its effectiveness in predicting the welder's responses.