Study On The Adaptive Control Method Of Maglev Artificial Heart Based On BP Neural Network

Heart disease has become a major threat to people's health,artificial heart transplantation is an important method to treat patients with severe heart disease,which has an important role and position in medical care,however there are very few heart donors,the need for artificial hearts is urgent,so the paper has important practical significance and academic value.To reduce the volume,simplify the working mechanism,and improve the efficiency,the paper proposes a new maglev artificial heart structure.Rotor friction of rotation process and Coupling between different magnetic fields is solved by using separated suspension structure and rotating structure.On the basis of parameters'value,the permanent magnetic force,electromagnetic suspension force and electromagnetic torque are deeply analyzed,the mathematical model expressions are got,which are validated by Maxwell 3D simulation.Design the high reliability control system of maglev artificial heart.Adopt the control method of double closed loop,where inner loop regulates the suspension stability of maglev rotor and outer loop regulates the torque fluctuation of maglev rotor.Modularize each functional circuit,and transplant BP neural network code into CPU S3C6410,All these lay firmly hardware foundation to realize the neural network adaptive control.Use a 2-3-1 cell structure of neural network,train the electromagnetic suspension sample data,and get total 13 neural network initial parameters,including 6 hidden layer weights,3 output layer weights,3 hidden layer biases,1 output layer bias.Establish the BP neural network simulation model of electromagnetic susupension,analyze the step response characteristics and the weight variation under closed-loop disturbance,analyze the suspension stability rule of BP neural network under different external force,cross-sectional area,static current,power amplifier gain,sensor gain,and parameters with same variable rate.Use 2-4-1 cell structure of neural network,train electromagnetic torque sample data and get total 17 neural network initial parameters,including 8 hidden layer weights,4 output layer weights,4 hidden layer biases,1 output layer bias.Establish BP neural network simulation model of electromagnetic torque,analyze torque fluctuation,weight variation under closed-loop disturbance,analyze torque fluctuation rule of BP neural network under different external torque,structure size,rotation speed,power amplifier gain,sensor gain,parameters with same variable rate.