Research on APSO-WNN and its Application in Vibration Fault Diagnosis of Hydroelectric Generating Units [APSO-WNN及其在水電機組振動故障診斷中的應用研究]

作者：Xu Y.-C.; Xia H.-T.; Fang S.-C.; Lu M.

来源：Journal of the Chinese Society of Mechanical Engineers, 2021, 42(2): 163-172.

• Adaptive particle swarm optimization
• Fault diagnosis
• Hydroelectric generating units
• Wavelet neural network

Based on the research of wavelet neural network (WNN), an adaptive particle swarm optimization (APSO) is proposed to solve the complex nonlinear relationship between the vibration characteristics and fault types of hydropower systems. This algorithm combines characteristics of evolutionary compute and swarm intelligences. It can change inertia weight according to the states of the particle adaptively. APSO rises the training speed of wavelet neural network, and improves network training accuracy. Experiments indicate that wavelet neural network based on APSO contains a higher precision and faster speed of diagnosis, compared with back propagation (BP) neural network and wavelet neural network. The algorithm is a new method for fault diagnosis of hydroelectric generating units (HGU), and it can be effectively applied to practical engineering.

• 单位
  三峡大学