Fault Prognosis Based on Restricted Boltzmann Machine and Data Label for Switching Power Amplifiers
Jinyong Yao, Tao Sheng, Zhen Jiangyun, Xiaolong Bao · 2018
The high efficiency and small size of the switching power amplifier (SPA) make it more ideal than amplifiers of other types, and it has been utilized widely. Effective fault prognosis of the SPA is extremely necessary for improving system reliability. This paper proposes a way to use these state variables containing a large number of component fault information to predict system faults. This method relies on a Restricted Boltzmann Machine (RBM) and its derived algorithm, which has excellent information filtering, component analysis and feature extraction capabilities. The classification-restricted Boltzmann model was constructed, and its classification performance was tested using data sets. It has good performance in avoiding over-fitting and local optimal solutions, and reducing machine learning process complexity.