Abstract: In order to investigate the resistance spot welding process for dissimilar steel sheet with unequal thickness,nonlinear multiple orthogonal regression assembling was applied to design the experiment.By taking nugget diameter and nugget deviation of spot welds as the study indexes and welding parameters such as welding current,electrode force,welding current duration and heat-treatment pulse current,and interactions among them as the influencing factor,a nonlinear multiple regression model of nugget geometry parameters was developed.The results showed that there was an effective prediction on nugget size by the optimized models.With these prediction results,the welding process was also optimized based on the analysis on the effects of parameters and their interactions on the welding quality.