In order to solve the problem that adaptive differential evolution algorithm based on reflection mutation strategy is prone to local optimization, a differential evolution algorithm based on mixed mutation strategy is proposed by introducing a basic mutation strategy, DE/current-to-rand/1. In order to improve the global convergence of the algorithm, the percentage of successful subgenerations based on the mutation strategy uses the roulette selection to select the appropriate mutation strategy for each individual. In this paper, the proposed algorithm combined with the finite element stress field is applied to the critical sliding surface search and the safety coefficient solution of two classical calculation examples. The results are compared with other limit equilibrium methods, and one example is used as a computational model to compare the convergence performance with other optimization algorithms. The results show that the improved algorithm proposed here has more stable performance and faster convergence speed. It also proves the effectiveness of the algorithm in combination with the finite element stress field to solve the slope problem.