Fault Diagnosis Using Imbalanced Data of Rolling Bearings Based on a Deep Migration Model

Abstract

To address the problem that uneven sample distribution can affect the accuracy and stability of fault diagnosis outcomes, we propose a deep transfer learning-Res2Net-convolutional block attention mechanism model. Firstly, the deep migration technique is used to transfer weights of the imbalanced source domain samples to the balanced target domain, expanding the data samples. Secondly, in the feature extraction and detection phase, All eight residual blocks are embedded in convolution block attention to ensure that interference signals are suppressed and that the key fault features are retained. Multilayer feature fusion extracts faulty sample features from multiple residual blocks of the network, which are combined into the feature fusion layer by parallel splicing. Finally, this model is experimentally validated using two different bearing datasets, and the combined evaluation indexes of the new model under the severe imbalance condition are 95.80% and 95.85%, respectively, demonstrating the feasibility and excellent performance of the model.

Detection and Classification of Rolling Bearing Defects Using Direct Signal Processing with Deep Convolutional Neural Network

“…The most widely used deep learning network structures are autoencoders [43], long short-term memory (LSTM) [7], and convolutional neural networks (CNNs) [36,[44][45][46]. Compared to the shallow structures mentioned above, CNNs show higher accuracy when operating directly on the diagnostic signal presented in the form of multidimensional arrays or vectors [38,45], further reducing the time of the diagnostic process [36,37]. CNNs also provide automatic symptom extraction, reducing the role of an expert in the diagnostic process.…”

“…Due to the application of ready-made neural structures with an excessive number of neural connections presented in the literature, this article presents the optimization of the CNN structure to maximize the accuracy of performance and minimize the required time and computational resources. Due to the crucial role played in modern diagnostic systems by the immunity of the solution to external disturbances [45] and the stability of the system's operation, it is important to optimize the neural structure that is a key element of the diagnostic application [41]. A reduced number of network parameters translates directly into the ability to implement the solution and response time to an emerging defect.…”