Mechanical Resonance Suppression based on Self-tuning Notch Filter for Servo System

To effectively suppress mechanical resonance, the notch filter after the speed regulator is commonly used to control the torque current, so the parameters of notch filter, such as frequency, depth and width, will directly affect the control effect. Meanwhile, due to the influence of complex operating conditions, the off-line tuning notch filter cannot be matched with the real-time operating conditions, it is necessary to self-tune the parameters of notch filter to ensure that servo system remains satisfactory control performance. In view of the fact that the adaptive notch filter only considers the notch frequency and heavily relies on the accuracy and rapidity of frequency detection, an adaptive notch filter using an improved BP neural network for servo system is proposed in this paper. Compared with the traditional BP neural network, the momentum term is added to prevent the algorithm from falling into the local minimum, the adaptive learning rate is used to solve the shortcomings of slow convergence and long training time, and the activation function considering the feature of the notch resonance is added to improve the stability of the system. Through the simulation, the proposed method can quickly adjust the parameters to achieve the satisfactory effect, and can also effectively suppress the resonance when the resonant frequency changes in the running operation.