Abstract: The optimization of residual stress on the surface of face gear grinding is a complex nonlinear problem, and traditional optimization algorithms cannot achieve efficient and accurate solutions. Therefore, this paper proposes an intelligent optimization method for residual stress on the surface of face gear driven by experimental data of tooth surface grinding. By designing a series of experiments to obtain the tooth surface residual stress data set, a response surface model driven by basic data was established, and intelligent particle swarm multi-objective optimization with tooth surface residual stress and grinding efficiency as the optimization goals was realized, solving the problem of face gear surface Residual stress optimization problem. Experiments show that the relative error of this method ranges from 0.89% to 1.61%, which proves the effectiveness of the residual stress optimization method for surface gear generation grinding. The analysis found that the residual compressive stress on the tooth surface is positively correlated with the grinding depth and workpiece speed, and negatively correlated with the grinding wheel speed and grinding wheel feed angle. From the sensitivity analysis, the sensitivity of residual stress on grinding depth is the highest, followed by the grinding wheel feed angle, while the sensitivity of residual stress on grinding wheel speed and workpiece speed is relatively small.