Aiming at the problem of mechanical parameters that affect the real-time changes of permanent magnet synchronous motors affecting the performance of the servo system, in this paper, we propose a permanent magnet synchronous motor mechanical parameter recognition solution based on the radial basis function neural network. Based on the designed radial neural network, we conduct parameter recognition and analysis of the three major mechanical parameters of rotation inertia, viscous damping, and load torque, and verify the superiority of this solution through a comparison experiment. The result proves that compared to the forgetting factor recursive least-squares, the parameter recognition scheme based on the radial basis function neural network designed in this article is reduced by 44.4%, the recognition accuracy has increased by 8%, and can recognize the friction coefficient and load torque, so it has certain practical application significance.