Aiming at strong time-varying and non-linear characteristics in the process of ultrasonic motor operation

a motor speed proportion integration differentiation neural network（ PIDNN） controller is introduced. The controller network fuses the state transition function

and uses the PID control law prior knowledge to determine the initial value of the weights. Through the variable learning speed BP（ VLBP） algorithm

training results are quickly converged and it is difficult to fall into the local minimum

so as to stably and fastly control ultrasonic motor. Use parametric-variable motor model to simulate and verify controller. The simulation results show that the method has a fast response speed and high robustness

and can compensate the nonlinearity of the motor during operation by online learning.