Super welder in augmented reality welder training system: A predictive control approach

Skills needed for critical manual welding operations typically require a long time to develop, and shortage of skilled welders has thus become an urgent issue the manufacturing industry is currently facing. This paper proposes an innovative augmented reality welder training system to help accelerate the training process of the unskilled welder. In this teleoperated system, unskilled welder perceives weld pool image with an auxiliary visual signal (arrow with direction and amplitude) superimposed upon, and make speed adjustment. As the first study of this kind, this paper aims to establish a machine algorithm calculating the optimal welding speed for unskilled welder to follow, referred to as super welder. In particular, dynamic experiments are conducted in order to correlate the welding speed to the fluctuating 3D weld pool surface characterized by its width, length, and convexity. Moving Average (MA) model is firstly identified and an Auto Regressive Moving Average (ARMA) model is then proposed to improve the modeling performance. A model-based predictive control (MPC) algorithm is proposed to derive an analytical solution to determine the optimal welding speed, and simulation is performed to verify the effectiveness of the proposed control algorithm. To further demonstrate the super welder's performance, automated welding experiments are conducted. Results verified that the proposed controller is able to track varying set-point and is robust against different welding currents.