DC Motor Speed Measurement Technology Based On Kernel Ridge Regression And Its Implementation

DC motor has been widely used in industry because of its own technical advantages and excellent performance.DC motor generally uses speed feedback to achieve closed-loop control,which requires speed detection.The position sensor method needs to install the mechanical sensor on the motor stator,resulting in the overall volume of the motor and the system cost increase,and the reliability of the speed detection is easy to be affected by the working environment.Sensorless technique is used to estimate the speed and position of the motor by measuring the current and voltage of the motor,so as to avoid the problems caused by the mechanical sensor.The modeling accuracy of the motor model method is affected by the parameter change,which leads to the error of the speed estimation.The pulse detection method of Brushed DC motor and the back EMF method of Brushless DC motor are used to estimate the rotor position through the waveform of current or voltage without dependent on the accurate mathematical model.In this paper,the brushed DC motor and brushless DC motor are used as the object,and the speed detection method based on this sensorless method is studied:Firstly,taking the brushed DC motor as the research object.The pulse detection method is used to calculate the current frequency by detecting the zero of the current ripple according to the principle that the commutation current ripple of the brushed DC motor is proportional to the speed.The noise component in the current has an effect on the waveform,which leads to some pulse amplitude too low or no zero point,which is easy to be missed,which makes it impossible to measure the velocity accurately.To solve this problem,a current classifier is designed to identify the vertex and non-vertex,and the accurate rotor position is obtained.A number of features are designed to monitor the trend of current waveform,which are affected by noise.In the case of limited number of samples and noise pollution,the kernel ridge regression method is used to train the classifier,which improves generalization ability and nonlinear modeling accuracy.Then,taking the brushless DC motor as the research object,the detection noise and burst noise in the end voltage signal of the brushless DC motor lead to spurious zero crossing of the back electromotive force,which makes the rotor position be detected wrong.To solve this problem,an inverse electromotive force classifier is designed to discriminate zero and non-zero points,and accurate commutation time is obtained.Aiming at the nonlinear strong interference problem of the back EMF classifier model,the kernel ridge regression method is adopted to solve the classifier model,and the modeling accuracy is improved.Finally,the experiment platform of brushed DC motor based on TMS320F28335DSP is designed,and the control algorithm of the brushed DC motor is programmed by using CCS3.3 as the software development tool.The classifier parameters are obtained by off-line training of Matalb.The experimental results show that the speed detection method has higher accuracy.The experiment platform of BLDC motor based on TMS320F28335DSP is designed,and the program is programmed to verify the BLDC motor control algorithm.The experimental results show that this method has good effect on rotor position detection of BLDC motor.At the end of the paper,the research work is summarized,and the further research direction of this subject is put forward.