Fault Diagnosis Based On Neural Network

The condition monitoring and fault diagnosis of mechanical equipment plays a vital role in ensuring the safe operation of the equipment.In this paper,a fault diagnosis algorithm based on convolutional neural network is proposed.Convolutional neural network has good recognition performance on image processing,in view of this feature of convolutional neural network,the vibration signals generated by rotating machines are transformed into time-frequency graphs by time-frequency analysis.Firstly,two time-frequency methods are introduced,including Short Time Fourier transform and Continuous Wavelet transform.Then,the rolling bearing signal is processed.Finally,the characteristics of different time-frequency graphs are compared.In this paper,a reasonable convolutional neural network is constructed,and the gradient descent method is used to update the network parameters.The time-frequency graphs obtained from the time-frequency analysis are used as the input of the network.By comparing the classification results of the network under different time-frequency graph inputs,the classification effect obtained by using the wavelet time-frequency graph as the inputs of the network is better.The convolutional neural network can extract the features contained in the image,but it is not a good classifier.Therefore,we propose a hybrid structure.The convolutional neural network is used to extract the features,and the extreme learning machines as a classifier,and the hybrid structure is applied to bearing fault recognition to obtain a good recognition effect.It is also applied to the fault recognition of wind turbine to further verify the hybrid structure.