Due to the load characteristics of the motor and other nonlinear factors, it is difficult for traditional PID to obtain satisfactory control effect. A motor neural network control algorithm was put forward. The network was trained with machine learning algorithm and neural network model obtain certain adaptive ability. The gradient descent back propagation algorithm is used to obtain the optimal control parameters of the motor. Simulation and experimental results show that it can effectively suppress overshoot, shorten adjustment time, improve system response speed and overcome the uncertainty of controlled object and the influence of random disturbance. The control effect was significantly improved.