Fitting of Dynamic Characters for Magnetic Bearing Control System by Using Deep Neural Network

Magnetic Bearing Control System（MBCS）has two major aporias ， non-linear and strong coupling. Most of the traditional methods are severely weakened by these two problems. This paper presents a noval control algorithm to optimize the performance of Magnetic Bearing Control System 。The proposed method estimates the non-linear dynamic characters and decouples system by using a deep neural network （ DNN ） .The inputs of DNN are system PID controller parameters and some other dynamic coefficients ， outputs are Magnetic Bearing speed，torque and power output characters. We utilize root-mean-square error（RMSE）as the loss function of DNN since the network is a regression model . Experiments showed that the proposed method could affectively fit and decouple dynamic characters for MBCS and performs competitively against Humans.