Performing non-linear anomaly detection analysis using Renyi entropy and ISSA-SVM

Abstract In industrial systems,the signal of rotating machinery is usually non-stationary, non-linear, and with noise interference.To improve the accuracy of anomaly detection analysis and overcome the limitations of optimization methods, This article proposes a rolling bearing fault diagnosis method using Renyi entropy and the integrated sparrow search algorithm (ISSA) with flight strategy for optimizing support vector machines (SVM). Firstly, wavelet packet analysis is used to decompose the original signal, and the optimal frequency band is selected from the decomposed bands for reconstruction. The reconstructed frequency band is then used to calculate the Renyi entropy and form the feature vector, which is input into the sparrow search algorithm with dynamically reverse learning factors for fault diagnosis. This algorithm improves the diversity of the population and the problem of easily getting stuck in local optima of the sparrow search algorithm by initializing the population with a flight strategy and adjusting the step size factor. The improved algorithm is compared with the diagnostic results of grey wolf optimization algorithm, sparrow search algorithm, and particle swarm optimization algorithm, and it is evident that the ISSA-SVM with improved algorithm has faster convergence and higher accuracy.