23 February 2021 in IEEE Sensors Journal
Bearing fault diagnosis is essential for the safe and stable operation of rotating machinery. Existing methods use signals from a single dimension, limiting diagnostic generality and accuracy. To address these limitations and make improved use of signal features from multiple dimensions, a novel convolutional neural network model with multi-dimensional signal inputs and multi-dimensional task outputs called MIMTNet is proposed. First, frequency domain signals and a time frequency graph are obtained by using the short-time Fourier transform and a wavelet transform to process original time domain signals simultaneously. Then, the time domain signals, the frequency domain signals, and the time frequency graph are fed into the model and a special aggregation is performed after extracting features from the three corresponding branches. Finally, the outputs of the three-dimensional tasks are acquired by different full connection layers to process the aggregated features of bearing position, damage location within the bearing, and the damage size. Two common bearing vibration signal datasets are used to verify the generalization ability of our proposed method. And experimental results show that the proposed method effectively improves the bearing diagnosis capability of the deep learning model.
1 September 2021 in IEEE Transactions on Instrumentation and Measurement
Fault diagnosis has a direct impact on the economic benefits of the modern rotating machinery industry. The input of current fault diagnosis methods is mostly the original vibration signal or the time-frequency graph (TFG) obtained by short-time Fourier transform. Both of these inputs use many data points, which reduce the real-time performance of the fault diagnosis model. Therefore, a novel fault diagnosis model is proposed to achieve high accuracy when the input data points are few. First, the two inputs of the model are the frequency-domain signal (FDS) and TFG obtained by processing the original signal with fast Fourier transform and continuous wavelet transform. Then, convolution (Conv) and deformable atrous Conv are used to extract the FDS and TFG features, respectively. These features are then fused by squeezing-and-excitation aggregation. Finally, the outputs of three different dimensions are obtained. Experimental results show that the proposed feature extraction and fusion module can increase the generality of the diagnostic model, and the proposed model has a better effect on three outputs compared to state-of-the-art methods.