Fault Diagnosis Of Off Design Condition Rolling Bearing Based On 1DCNN

As an important part of rotating machinery,rolling bearing is closely related to the safety and stability of the whole mechanical equipment.Therefore,real-time detection of the running state of rolling bearing and timely understanding of the fault location and severity can effectively reduce economic losses and ensure the safety of production personnel.However,in the actual production process,the working load of the rolling bearing in the working state often changes,which leads to the great difference between the distribution of the training data of the rolling bearing fault diagnosis model and the test data in the actual production process.Therefore,in order to reduce the distribution difference between the training data set and the test data set,on the basis of domestic and foreign scholars’ research and traditional rolling bearing fault diagnosis,this paper proposes a method based on 1dcnn to solve the rolling bearing fault diagnosis under variable working conditions:(1)Aiming at the problem that the traditional fault diagnosis method needs a lot of manpower cost and time,the one-dimensional convolution neural network is used for fault diagnosis of rolling bearing.Firstly,the original vibration signal collected by the sensor is preprocessed by fast Fourier transform,and the preprocessed data is used as the input of one-dimensional convolution neural network for training.The trained model is used for adaptive feature extraction and classification of vibration signals with different fault degrees,and finally the judgment results of fault diagnosis are obtained.The experimental results show that compared with other traditional fault diagnosis methods,the one-dimensional convolutional neural network has higher diagnostic accuracy.(2)Aiming at the problem of low accuracy of fault diagnosis caused by the large difference of original vibration signal distribution under different working conditions,a rolling bearing fault diagnosis method based on one-dimensional convolution neural network with limited batch normalization is proposed.The original vibration signal collected by the sensor is preprocessed by fast Fourier transform,and the data obtained is used as the input of neural network for training.Through the trained model,the vibration signal under different working conditions is extracted and classified adaptively,and the restricted batch normalization is applied to the full connection layer to reduce the difference of signal distribution under different working conditions,so as to improve the performance of the model The off design fault diagnosis performance of vibration signal.Experimental results show that: compared with other typical fault diagnosis methods,the proposed method has higher diagnostic accuracy,better stability and generalization performance under off design conditions.(3)In rolling bearing fault diagnosis,the original vibration signal information collected by single sensor is incomplete,which leads to the low accuracy of off design fault diagnosis.In view of this,this chapter proposes a multi-sensor rolling bearing fault diagnosis method based on one-dimensional convolution and dynamic routing algorithm.Firstly,the time-domain vibration signals collected by multiple sensors are converted into frequency-domain spectrum by fast Fourier transform;secondly,the multi convolution network is constructed to extract features from the amplitude of the spectrum and normalize them;then the dynamic routing algorithm is used to fuse the features extracted by the multi convolution network to train the model;finally,the vibration data under variable conditions are imported into the model to realize fault diagnosis.Experimental results show that: compared with other typical intelligent fault diagnosis methods,the method proposed in this chapter has higher diagnosis accuracy,better stability and generalization performance under off design conditions.