A High-Accuracy Fault Detection Method Using Swarm Intelligence Optimization Entropy

Abstract

Entropy theories play a significant role in rotating machinery fault detection. The key parameters of these methods are, however, often selected subjectively based on trial-and-error methods or engineering experience. Unsuitable parameters would result in an inconsistency between the extracted entropy results and the realistic case. In order to address this issue, a complexity measurement method called “swarm intelligence optimization entropy” (SIOE) is proposed, which adaptively estimates optimal parameters using skewness metrics, logistic chaos theory, and African vulture optimization (AVO). By considering the variability and dynamic changes of various signals, SIOE enables the extraction of robust and discriminative dynamic features. Additionally, a collaborative intelligent fault detection method for rotating machinery fault detection is developed, based on SIOE and extreme gradient boosting (XGBoost). This method aims to accurately identify single faults, compound faults, and varying fault degrees within the rotating machinery. Simulation and fault detection experiments on rotating machines demonstrate that SIOE improves recognition accuracy by up to 21.25% compared to existing entropy methods. The proposed intelligent fault detection method improves recognition accuracy by up to 15.71% compared to advanced fault detection methods. These results highlight the advantages of SIOE in complexity measurement and feature extraction, as well as the effectiveness and accuracy of the proposed intelligent fault detection method, in identifying rotating machinery faults.