Nonlinear identification and self-learning CMAC neural network based control system of laser welding process

Laser welding process plays a critical role in modern processing technologies. It is a typical nonlinear system which is difficult to model and control. In this paper, a laser welding system in University Kentucky is studied. At first, a SISO nonlinear discrete Hammerstein model is established for this system, the welding speed is selected as input and the welding width is output. This model is identified using least-square iterative identification algorithm and validated by simulation experiment of step response. Afterwards, a self-learning CMAC neural network based control system is designed for the laser welding process. To improve the performance, a PID controller is attached to it. Finally, this control system is confirmed effectively by four simulation experiments. Results indicate that the identified model and control system are practicable in real systems.