A Principal Components Rearrangement Method for Feature Representation and Its Application to the Fault Diagnosis of CHMI

Abstract

Cascaded H-bridge Multilevel Inverter (CHMI) is widely used in industrial applications thanks to its many advantages. However, the reliability of a CHMI is decreased with the increase of its levels. Fault diagnosis techniques play a key role in ensuring the reliability of a CHMI. The performance of a fault diagnosis method depends on the characteristics of the extracted features. In practice, some extracted features may be very similar to ensure a good diagnosis performance at some H-bridges of CHMI. The situation becomes even worse in the presence of noise. To fix these problems, in this paper, signal denoising and data preprocessing techniques are firstly developed. Then, a Principal Components Rearrangement method (PCR) is proposed to represent the different features sufficiently distinct from each other. Finally, a PCR-based fault diagnosis strategy is designed. The performance of the proposed strategy is compared with other fault diagnosis strategies, based on a 7-level CHMI hardware platform.

Rolling Bearing Fault Diagnosis Based on an Improved HTT Transform

“…Principal component analysis has shown its effectiveness in extracting main information from high-dimension data [ 39 , 40 ]. It is used for matrix de-noising in this paper.…”