Rolling Bearing Fault Diagnosis Algorithm Based on RCMSDE Fusion Transfer Learning

To address the issue of poor accuracy in cross- condition diagnosis with traditional data-driven fault diagnosis methods, a bearing fault diagnosis method based on refined composite multiscale symbolic dynamic entropy integrated with transfer learning and least squares support vector machine is proposed. The refined composite multiscale symbolic dynamic entropy is used to extract fault features, combined with balanced distribution adaptation for feature mapping, and Bayesian optimization is applied for parameter tuning. The mapped feature vectors, with optimized parameters, are then input into an LSSVM optimized by the improved grey wolf algorithm to obtain the final fault identification accuracy. The proposed method was applied for cross-condition fault diagnosis experiments on the Case Western Reserve University dataset and the Paderborn University dataset, construct the test dataset using data with multiple fault types and varying severity levels. Experimental results show that the migration experiments between different working conditions on the two datasets achieved an average accuracy of 97% and 87%, respectively, and the highest accuracy rate can reach 100<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, It shows that the proposed method can accurately identify the fault type, and has a higher average accuracy than the comparison algorithm.