To deal with uncertainties in electrical engineering, a multi-objective reliability-based robust design optimization framework based on an adaptive learning two-stage surrogate model is studied. The suggested algorithm aims at enhancing both robustness and reliability of electrical equipment. The adaptive support vector machine model is introduced into the modeling of constraint boundaries in the first stage, and then the Kriging model improved by gradient-assisted learning strategy is used to model the target response in the safety domain, and finally the NSGA-II algorithm is utilized to obtain the uniformly distributed Pareto front. In this paper, this method is applied to the optimization design of permanent magnet synchronous motor considering reliability and robustness.