Sensorless Control Technology Study Of Electromagnetic Linear Driving Device

As a direct drive component,the electromagnetic linear driving device can convert the input electric energy into mechanical energy to drive linear motion,which is suitable for the linear motion of stable feeding,which has attracted wide attention.During the operation of the electromagnetic linear driving device,the displacement of the mover needs to be used as a feedback signal to control the progress of the motion.At present,the detection of displacement signals is basically realized by high-speed displacement sensors,which affects the cost and compactness of the system.Therefore,the realization of the position sensorless control technology has important significance for the development of linear drive control technology.This paper combines theoretical analysis,mathematical modeling,simulation research and experimental verification.Based on the position sensorless control technology of electromagnetic linear driving device,the direct calculation method,artificial neural network algorithm and improved GA-BP algorithm are proposed.The sample data obtained by the measurement and control system is used for training verification and analysis.The specific work done in this paper includes the following aspects:(1)The structure and principle analysis of the electromagnetic linear driving device is completed.The measurement and control system is established based on the direct drive AMT shift test bench,and the test procedure of the measurement and control system is determined.The data samples of the system parameters under various working conditions are obtained,which lays a foundation for the subsequent mover displacement estimation scheme.(2)The mathematical model of each subsystem of the electromagnetic linear driving device is established,and the mathematical relationship between the parameters of each subsystem is determined by mathematical model.The mover displacement estimation scheme based on the back electromotive force of electromagnetic linear driving device is established.The simulation model of MATLAB/Simulink is built and the simulation results obtained from the application sample data were analyzed.Due to the error of some system parameters of the electromagnetic linear driving device and the influence of the control strategy in the measurement and control system,there is a certain error in the first stage of the up shift stage and the down shift stage,but the displacement calculation error of the remaining retreat stage is gradually increased.The maximum absolute error at the beginning of the third up shift stage reached 6.05 mm.The displacement of the mover of the electromagnetic linear driving device obtained by the direct calculation method has certain error and is unavoidable,and cannot be used as the displacement feedback signal of the control system.(3)The principle of indirect displacement detection of electromagnetic linear driving device is established.The principle of RBF and BP neural network are analyzed respectively.The MATLAB network structure of two neural network algorithms is established,and the relevant network parameters are determined.The two kinds of neural networks were offline trained and verified by the sample data.The results show that under the experimental conditions of working condition 1,compared with RBF neural network,the prediction accuracy of BP neural network is improved by 2.04%,RMSE is reduced by 30.45%,MAPE is reduced by 25%,and training time is saved by 73.05%.Under the experimental conditions of working condition 2,compared with the RBF neural network,the prediction accuracy of the BP neural network is improved by 0.7%,the RMSE is reduced by 21.06%,the MAPE is reduced by 4%,and the training time is saved by 70.39%.Obviously,the prediction effect of BP neural network is relatively better,and the network structure is simpler.(4)In order to improve the accuracy of BP neural network prediction,improve the prediction effect of the down shift stage and the first up shift stage,a genetic algorithm with global convergence and a BP neural network algorithm with good fast local optimization performance are combined to construct an improved GA-BP algorithm.The MATLAB model of the algorithm is established,and the network is trained and verified by the sample data of each working condition.The obtained training results are analyzed: under various working conditions,the prediction accuracy of improved GA-BP algorithm is greater than 96.3%,which is 2.2% higher than that of traditional BP algorithm,and the RMSE is reduced by 21.8% on average,the MAPE is reduced by 23.9% on average.The measurement and control system of the electromagnetic linear driving device based on the direct drive AMT shift test bench is used to carry out the position sensorless control test: the trained improved GA-BP network is loaded into the measurement and control system.Taking the real-time parameters of the system as the input of the network,the predicted displacement of the network output replaces the output displacement of the original displacement sensor as feedback of the control system.The results of the position sensorless control test show that the improved GA-BP network can achieve smooth shifting,and the accuracy of the test results of each working condition is greater than 96%.