A rolling bearing fault diagnosis technique based on recurrence quantification analysis and Bayesian optimization SVM

Wang, Bing; Qiu, Wentao; Hu, Xiong; Wang, Wei

doi:10.1016/j.asoc.2024.111506

Cited by 11 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sun et al proposed a CNN-LSTM based model for bearing fault diagnosis in complex operating environments, and the results show that the model has better load generalisation capability and noise immunity [8]. Wang et al proposed the RQA-Bayes-SVM for the healthy diagnosis of bearings; experiments showed that RQA-Bayes-SVM has better performance in fault mode diagnosis and fault degree differentiation [9]. Zhao et al proposed the DenseNet-BLSTM for the problem of extracting features effectively using traditional fault diagnosis methods rolling bearings; experiments show that the DenseNet-BLSTM has good fault diagnosis capability [10].…”

Section: Introductionmentioning

confidence: 99%

Intelligent Fault Diagnosis of Rolling Bearings Based on Markov Transition Field and Mixed Attention Residual Network

Tong,

Zhang,

Wang

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

To address the problems of existing methods that struggle to effectively extract fault features and unstable model training using unbalanced data, this paper proposes a new fault diagnosis method for rolling bearings based on a Markov Transition Field (MTF) and Mixed Attention Residual Network (MARN). The acquired vibration signals are transformed into two-dimensional MTF feature images as network inputs to avoid the loss of the original signal information, while retaining the temporal correlation; then, the mixed attention mechanism is inserted into the residual structure to enhance the feature extraction capability, and finally, the network is trained and outputs diagnostic results. In order to validate the feasibility of the MARN, other popular deep learning (DL) methods are compared on balanced and unbalanced datasets divided by a CWRU fault bearing dataset, and the proposed method results in superior performance. Ultimately, the proposed method achieves an average recognition accuracy of 99.5% and 99.2% under the two categories of divided datasets, respectively.

show abstract

Section: Introductionmentioning

confidence: 99%