Scholarlink

<p>Voltage and current are commonly sampled signals for the diagnosis of permanent magnet synchronous motor (PMSM) demagnetization faults (DMF). Traditionally, external signal capture interfaces (ESCIs) have been a common choice; however, these devices have been hindered by noise interference, sampling errors, and high costs. This study pioneers the concept of ultra-low sampling frequency data (ULSD) and presents a novel approach to re-index and reconstruct voltage and current waveforms using the ULSD re-indexing method. Utilizing demagnetization flux linkage curves to derive back-electromotive force (Back-EMF) characteristics for establishing a demagnetization model. Crucially, the zero-cross offset (ZCO) is used to determine the period sampling points (PSPs), after which the data index is redefined. The waveform is then reconstructed based on a sampling period offset as a condition. Finally, the rearranged ULSD waveforms are integrated into a 1D CNN for practical DMF diagnosis. The primary contribution of this study is the effective reconstruction of Back-EMF waveforms using the ULSD re-indexing method. When applied to a 1D CNN for diagnosis, the data obtained from the ULSD re-indexing method achieve a recognition rate of 85.25%, outperforming the data from the ESCI method.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

IPMSM demagnetization fault diagnosis based on ultra-low sampling frequency data re-indexing restoration method

  • 期刊论文

  • 发表日期：

    2025年11月28日

查看全文
  • Chin-Sheng Chen,
  • Chia-Jen Lin,
  • Jia-Feng Liu,
  • Feng-Chieh Lin

摘要

Voltage and current are commonly sampled signals for the diagnosis of permanent magnet synchronous motor (PMSM) demagnetization faults (DMF). Traditionally, external signal capture interfaces (ESCIs) have been a common choice; however, these devices have been hindered by noise interference, sampling errors, and high costs. This study pioneers the concept of ultra-low sampling frequency data (ULSD) and presents a novel approach to re-index and reconstruct voltage and current waveforms using the ULSD re-indexing method. Utilizing demagnetization flux linkage curves to derive back-electromotive force (Back-EMF) characteristics for establishing a demagnetization model. Crucially, the zero-cross offset (ZCO) is used to determine the period sampling points (PSPs), after which the data index is redefined. The waveform is then reconstructed based on a sampling period offset as a condition. Finally, the rearranged ULSD waveforms are integrated into a 1D CNN for practical DMF diagnosis. The primary contribution of this study is the effective reconstruction of Back-EMF waveforms using the ULSD re-indexing method. When applied to a 1D CNN for diagnosis, the data obtained from the ULSD re-indexing method achieve a recognition rate of 85.25%, outperforming the data from the ESCI method.