Research On Mechanical Fault Diagnosis Method Of Conventional Circuit Breaker Based On Convolution Nueral Network

The conventional circuit breaker is the protection and control equipment in the low-voltage distribution system.Its health status has a huge impact on the performance and stability of the distribution system.The fault diagnosis can discover the hidden trouble and the trend of the fault in time,which is helpful for the maintenance personnel to grasp the health status of each component of the circuit breaker in time,formulate the maintenance strategy and achieve predictive maintenance.With the development of deep learning theory,the idea of learning fault features from data itself provides a new way for fault diagnosis of circuit breakers.As one of the most important models in deep learning,convolutional neural network has a strong ability of data representation,which makes it have a strong advantage in distinguishing fault types and other issues.It gets rid of the dependence of traditional fault diagnosis methods on artificial feature extraction and expert experience,and opens up a broad development prospect for the field of conventional circuit breaker fault diagnosis.Thus,this thesis does research on fault diagnosis technology for switching operation accessories and contact system of the conventional circuit breaker.Firstly,the basic structure and working principle of the conventional circuit breaker are described in detail,the common fault types and causes of the switching operation accessories and contact system are analyzed.And carry out simulation tests on different mechanical states,design and build a fault signal acquisition system.By analyzing and selecting the switching coil current and contact vibration signals as the fault characteristic signals of switching operation accessories and contact system respectively.In addition,according to the time-frequency analysis method of contact vibration signal,the continuous wavelet transform with higher time-frequency resolution is selected for time-frequency transformation of vibration signal.Then,since AC power supply is adopted in the coil circuit of switching accessories for low voltage conventional circuit breaker,the randomness of the closing phase angle of the coil circuit may cause the difference of current signals under the same operating state.an intelligent fault diagnosis algorithm based on adaptive one-dimensional deep convolutional neural network with wide first-layer kernel(AW-1DCNN)is proposed.Compared with the traditional intelligent fault diagnosis method including two stages of manual feature extraction and fault classification,the proposed method combines these two stages into one.The original coil current signal is used as the input of the AW-1DCNN model to train the model and maximize the advantages of the model to automatically learn the characteristics of the original signal,can effectively overcome the influence of closing phase angle on fault diagnosis results and improve the accuracy of fault diagnosis.Finally,in order to eliminate the pre-defined influence of parameters caused by manual feature extraction and improve the diagnosis effect at the same time,in this thesis,according to the common faults of contact system,the vibration signal is transformed into time-frequency image as the research object,a two-dimensional convolutional neural network fault diagnosis method for fault diagnosis of contact system is proposed.This method combines the time-frequency analysis method of continuous wavelet transform with the improved alexnet convolution neural network model to realize the fault diagnosis of contact system.In order to improve the fault diagnosis recognition rate and stability of the model,the Alex Net network structure,parameters,and optimization algorithms were improved.Experimental results show that the improved Alex Net network model has higher fault recognition rate,and the network model is more stable.