Abstract: To control weld penetration depth in real time in pulsed gas mental arc welding (GMAW-P) process, the voltage variation amplitude during a peak current period (ΔU) is proposed to characterize the change in weld penetration depth, and ΔU is measured and controlled to control weld penetration depth in GMAW-P process. A single input and single output system is established with ΔU as system output and base current as system input for the weld penetration depth control. The static model between the input and output of the system shows that the weld penetration depth control system has nonlinearity, which can be described by the Hammerstein model with the interference. The recursive least square method is added in classical predictive control algorithm, which is established based on the Hammerstein model, to identify the system model parameters online and to achieve adaptive control of the weld penetration depth. The control algorithm simulation results and the real time welding experiments prove that the adaptive predictive control algorithm can well control the weld penetration depth in GMAW-P process. The effectiveness and adaptability of the control algorithm are verified by variable heat dissipation experiments.

Key words: Hammerstein model, adaptive predictive control, simulation, weld penetration depth control, GMAW-P