1. State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources (North China Electric Power University), Beijing 102206, China;2. Test and Research Institute of China Datang Corporation Renewable Power Co., Limited, Beijing 100068, China
Abstract:Fault diagnosis of rolling bearings is the key to improve equipment availability and reduce operation and maintenance cost. Least square support vector regression (LSSVR) is a kind of effective method for fault diagnosis, and glowworm swarm optimization (GSO) algorithm is applied to obtain the optimal combination of penalty parameter and kernel parameter which are always restricted by subjective experience. A rolling bearing fault diagnosis method based on LSSVR optimized by GSO was proposed. Experiments show that the presented method can precisely diagnose both fault location and fault severity of rolling bearings, with higher accuracy compared with normal LSSVR and BP neural network, which has validated the reliability of the proposed method.