Nonlinear Identification of Laser Welding Process

It has been well recognized that weld pool geometry plays a critical role in fusion welding process such as laser welding. In this study, the authors establish a standard diode laser welding system and perform a series of experiments to investigate correlations between welding parameters and the weld pool geometry. Custom digital camera for image acquisition and software for image processing are implemented in the system to obtain the surface width of the weld pool. Experimental data has demonstrated significant nonlinearity in the diode laser welding process. The authors thus propose a continuous Hammerstein identification methodology to approximate this process. A single-input–single-output (SISO) nonlinear continuous model is then identified and validated for the diode laser welding process using the experimental data. Because by nature laser welding is a heat transferring process in which the heat applied to a unit length of the work-piece along the weld seam is inversely proportional to the travelling speed, the model takes the reciprocal of welding speed as the input and the top surface width as the output. For this methodology, all experiments need to be dynamic and be conducted using either Step or PRTS (pseudo-random ternary signal) inputs. In a revised and simplified version, the linear dynamics is identified first from a single dynamic experiment and then used in the identification of the nonlinearity with other static response experiments. The validation proves both identification methods are capable of predicting the surface width of the weld pool.