Demagnetization Fault Diagnosis of PMSM Using Custom Phase Space Reconstruction Image

In traditional one-dimensional (1-D) signal-based fault diagnosis methods, there are difficulties in signal feature extraction and complexities in adjusting model hyperparameters. Therefore, a demagnetization fault diagnosis method for permanent magnet synchronous motor (PMSM) using a custom phase space reconstruction (CPSR) image is proposed, which can be generalized to condition recognition. First, a CPSR method is proposed to reconstruct the magnetic leakage signal into a two-dimensional (2-D) data image. Second, to overcome the problem of poor diagnosability of high-dimensional shallow features in existing local feature extraction methods, an image feature coding (IFC) method is proposed, which can improve diagnostic performance through deep feature extraction and coding. Third, the improved autonomous learning multimodel constructs different fault data clouds with coded information, and realizes fault diagnosis according to the fuzzy rules of the data cloud. Four kinds of demagnetized fault motors were manufactured for experimental verification. The extensive experimental results demonstrate that the proposed method is effective.