For the shortcoming of slow convergent speed and no objective criterion of convergence criteria of the simple genetic algorithm,an improved algorithm which combined with the tapered simplexpartition ideology was proposed.Firstly,the optimal solution problem was transformed into a fixedpoint problem;then the solution space was spited into the simplex subdivision and perform iterativesearch on the basis of the vertices information of simplex;finally,the full labeled simplex was converted to the target value to output.This paper established the optimized design model which minimizedthe electromagnetic radial force and applied it to the structural optimization of switched reluctance motor.With a 2.2kw power for the prototype as an example,the author used the model to optimizationanalysis,and compared with the original motor parameters.The results show that the global optimalsolution is found after 10 iterations,which prove the improved algorithm is efficient and steady.