Abstract
The traditional bearing fault diagnosis methods have complex operation processes and poor generalization ability, while the diagnosis accuracy of the existing intelligent diagnosis methods needs to be further improved. Therefore, a novel fault diagnosis approach named CNN-BLSTM for bearing is presented based on convolutional neural network (CNN) and bidirectional long short-term memory (BLSTM) in this paper. This method directly takes the collected one-dimensional raw vibration signal as input and adaptively extracts the feature information through CNN. Then, the BLSTM is used to fuse the extracted features to acquire the failure information sufficiently and prevent the model from overfitting. Finally, two different experimental datasets are used to verify the effectiveness of the method. The experimental results show that the proposed CNN-BLSTM model can accurately diagnose the fault category of bearings. It has the advantages of rapidity, stability, antinoise, and strong generalization.
Highlights
As a commonly used power transmission device in mechanical equipment, bearings are widely used in industrial production [1,2,3]
In order to explore the influence of too many classifications on the diagnosis results, faults of different sizes at the same location were regarded as different classes. e data categories in Table 1 were divided into 4 and 10 categories for experimental comparison. e comparison model was the method proposed in this paper, the convolutional neural network (CNN)-LSTM model framework with bidirectional long short-term memory (BLSTM) layer replaced by LSTM layer and the CNN model framework without BLSTM layer
When signal-to-noise ratios (SNR) 0 dB, the CNN-BLSTM model could achieve 100% test accuracy
Summary
As a commonly used power transmission device in mechanical equipment, bearings are widely used in industrial production [1,2,3]. Levent et al [29] directly input the raw signal into one-dimensional CNN for fault classification and verified it on the Case Western Reserve University (CWRU) bearing dataset These methods can better extract fault features and perform fault diagnosis, the accuracy of diagnosis is not high, the operation process is complicated, and the running time is longer. (1) A new feature extraction method, called CNNBLSTM, is proposed to reduce the complexity of the diagnosis process and avoid overfitting of the model during the training process (2) Avoid the interference of high correlations between different size fault classes in the same position (3) is method has good robustness and achieves high accuracy even under strong noise interference e rest of this paper is as follows. Where w is the t-th activation width value of of the the pooling area; ytj(w) j-th neuron in the represents the l-th layer; ztj(l) represents the t-th value after the pooling operation in the j-th neuron in the l-th layer
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
