Study On Motor Fault Diagnosis Method Based On Multi-scale Convolutional Neural Network

In the process of modern industrial production equipment continuously moving towards the direction of structure,automation and intelligence,motors are still the main power output equipment.Failure of the motor during operation will cause problems such as reduced operating efficiency and increased system energy consumption.In severe cases,the motor will be damaged,causing the overall system equipment to be stopped for maintenance for a long time,causing serious economic losses.Therefore,the research of motor intelligent fault diagnosis technology is of great significance for ensuring the stability and reliability of efficient operation of production equipment.With the continuous innovation and development of science and technology,breakthroughs have been made in signal processing,artificial intelligence and other technologies,and fault diagnosis technology has become more precise and intelligent.The thesis combines the common fault diagnosis problems of the motor in the actual production process and the fault diagnosis in the strong noise environment.Based on the analysis of the failure mechanism,the intelligent fault diagnosis method of the motor is studied in depth.(1)Firstly,the intelligent fault diagnosis method based on signal processing is studied,the mechanism of motor fault generation based on vibration signals is explored.Based on the analysis of the motor fault characteristics under different operating states,the signal analysis method based on empirical mode decomposition is studied,and aimed at it Existing modal aliasing is proposed to use set empirical modal decomposition to analyze the motor vibration signals,select the first 4 orders of IMF components through correlation analysis and calculate the corresponding envelope spectrum and marginal spectrum as the feature extracted signal sequence.(2)Calculate 9 different time and frequency domain statistical features in the multi-sequence sample signal to obtain the original feature set.Based on the analysis of the clustering algorithm,a sensitive feature based on adjusting mutual information and standard deviation is proposed.Select a method to filter features from the original feature set to construct a sensitive feature set for motor fault diagnosis.In addition,in view of the problems of feature interference and redundancy in the feature set,a feature dimension reduction method is proposed to reduce the feature dimension.Finally,two more popular machine learning algorithms,Support Vector Machines and Extreme Learning Machine,are used to implement motor fault diagnosis and verified by comparison experiments.(3)Aiming at the problems of complex process,relying on expert knowledge and insufficient learning ability of shallow structure features in traditional intelligent fault diagnosis methods based on signal processing,the end-to-end convolutional neural network algorithm is used for motor fault diagnosis.In order to extract the rich complementary features of different scales in the signal,a multi-scale fusion framework is proposed using three different size convolution kernels of 3 × 1,5 × 1and 7 × 1,and a multi-scale one-dimensional convolution neural network Motor fault diagnosis method.Finally,the superiority of the proposed method under variable operating conditions and noise interference is verified through experiments.(4)In order to improve the recognition efficiency and accuracy of the MS-1DCNN fault diagnosis method under variable motor conditions and strong noise interference,the residual network structure is introduced to deepen and improve the network,and the features are further enhanced on the basis of the multi-scale framework learning ability.The implementation principles of the two attention mechanism algorithms of the squeeze and excitation module and the convolutional attention module are studied respectively.The attention module suitable for the one-dimensional residual network is designed and embedded into the residual module to construct two Multi-scale attention residual network model.Finally,the validity and superiority of the proposed model are verified using experimental bench data.The article has 40 figures,18 tables and 92 references.