Bearing Fault Diagnosis Based on Frequency Domain Feature Extraction and Combined Dual-stream CNN
Aiming at the problem that various fault features of bearings extracted by traditional methods are mixed and some types of faults are hardly to be distinguished,a fault diagnosis model based on dual-stream convolution neural network(CNN)is designed.Firstly,the vibration signal is converted to the frequency domain.In order to reduce the interference of low frequency weak signal,the frequency domain signal is filtered and the passband and stopband attenuation values of the filter are determined to ensure that the signal is not distorted.Then,the frequency bandwidth is determined,and the range of the maximum signal amplitude ratio is obtained under this bandwidth value.At the same time,the average value of the normal signal amplitude is used as the threshold value of the high-frequency signal to determine the maximum frequency.The above parameters are used as the parameters of filters,and the signal is filtered to obtain the spectrum signal and construct the time-frequency image.The extracted spectrum signal and time-frequency map are used as the input of the two channels of the model.A feature fusion layer is added behind the convolution layer and the pooling layer.The features of the two channels are fused by weighted fusion,which significantly improves the discrimination of various fault features.A fault platform is built to collect data for verification.The experimental results show that the method can extract the unique features of each type of fault,and the accuracy of bearing fault identification reaches 98.95%.
fault diagnosisfeature extractionconvolution neural networkwave filtertime frequency diagramfeature fusion
万方数据
维普
