Identification Of Rotor Unbalance Of Micromotor Based On Deep Learning

Micro-motors are widely used in various fields.As the core rotating part of the motor,the rotor affects the safe and stable operation of the motor.Rotor unbalance caused by material defects,machining and assembly errors,and rotor structure asymmetry is one of the main causes of motor failure.Therefore,the rotor needs to be corrected by dynamic balance test before installing the motor rotor.In order to correct the unbalanced rotor efficiently and accurately,it is a key problem to accurately identify the unbalance of the rotor in the detection stage.Due to the light weight and small size,the unbalance of the micro-motor rotor is very small,and the dynamic balance accuracy is higher.However,the traditional identification methods of rotor unbalance have limitations and cannot meet the high-precision and high-quality balance requirements of micro-motor rotors.Therefore,it is more and more urgent to study new rotor unbalance detection methods.Aiming at the above problems,the research is carried out with the goal of identifying the unbalance of the rotor with high precision and high efficiency.The main research works are as follows:Aiming at the problem of low repeatability of rotor unbalance detection results,the basic theory of rotor dynamic balance,the classification of rotor unbalance,the detection method of dynamic balance and the correction principle of dynamic balance are studied,and it is pointed out that the influence coefficient matrix affects the detection results of rotor unbalance.The average influence coefficient method is proposed.After the influence coefficients at different weighting phases are obtained by weighting at the equal division of the rotor circumference,then they are averaged.By analyzing the condition number of the influence coefficient matrix,it is proved that the average influence coefficient method can avoid the maximum condition number of the influence coefficient matrix,improve the accuracy of the influence coefficient matrix,and make the detection result of the rotor unbalance more accurate.Through the test of three-groove and fivegroove rotors,the results show that the amplitude caused by the residual unbalance is reduced after the balance correction of the rotor by the proposed average influence coefficient method,which verifies the effectiveness of the average influence coefficient method.Targeted to the problem of inaccurate measurement of rotor vibration response,the dynamic model of rotor-bearing system is constructed according to the actual situation of rotor dynamic balance tester.The differential equation of motion of rotor-bearing system is obtained by force analysis of the model.Secondly,the Newmark-? numerical algorithm is used to solve the vibration response of the rotor.By comparing with the vibration response of the real test,the results show that the response obtained by the simulation is consistent with the response obtained by the real test,which verifies the accuracy of the model.The rotor vibration response obtained by the model simulation can be used as the input data for identifying the rotor unbalance based on the deep learning method in Chapter 4 to train the model.Then,in order to reduce the near-frequency interference caused by the natural frequency of the system on the vibration response signal,the influence of each design parameter on the inherent frequency of the system is studied and obtained,and the method of effectively reducing the inherent frequency of the system is proposed through simulation experiments.Because the vibration response of the rotor has a great influence on the final unbalance detection result,in order to measure the vibration response of the rotor more accurately and sensitively,the sensitivity equation of the rotor vibration response to each design parameter is derived by using the direct differential method,and the variation law of the response sensitivity is obtained.It is verified by experiments that the distance from the support to the rotor mass center O have a great influence on the rotor vibration response,which is regarded as the key design parameters and brought into the model training as the variables of the neural network input layer in Chapter 4.Due to the many limitations of the traditional rotor unbalance detection method,the balance correction effect is not ideal.An identification method of micro-motor rotor unbalance based on deep learning is proposed.Firstly,combined with the vibration response data obtained from the model simulation and the rotor unbalance calculated by the average influence coefficient method,the dataset is made and divided into training set,verification set and test set to prevent over-fitting during the training process.Then the data is normalized to improve the convergence speed and training accuracy of the neural network.Secondly,the neural network architecture is constructed by LSTM,including two LSTM layers,two hidden layers and a fully connected layer.According to the characteristics of input data and output data,the Re Lu function is selected as the activation function of the hidden layer,and the training parameters such as learning rate,batch size and training times are reasonably configured.Then,the input data is used to train the model and evaluate the model.Finally,the trained model is used to identify the rotor unbalance.The residual unbalance after balance is smaller than that of the average influence coefficient method,which verifies the effectiveness of the detection method.In addition,the identification method rotor unbalance based on deep learning only needs to measure a set of unbalanced vibration responses to identify the unbalance of the rotor through the neural network,and does not need to start and stop many times to obtain its influence coefficient,which has higher detection efficiency.