Scholarlink

<p>In safety–critical applications, preventing the complete failure of power converters is crucial. However, existing anomaly detection methods primarily focus on single-component faults, neglecting practical scenarios where multiple components degrade simultaneously due to various stressors. To address this limitation, this paper presents a systematic methodology for detecting and analyzing anomaly states resulting from multi-component degradation in power converters. Statistical parameters derived from the output voltage are used to identify anomalies, capturing the system's dynamic changes as faults progress from single to multiple components. A random forest regressor is employed to determine the most significant features contributing to fault detection. Additionally, a Fault Severity Index (FSI) is computed to provide a quantitative measure of degradation severity. The methodology is validated through LTspice simulations on three converter types: Single-Ended Primary Inductor Converter (SEPIC), Boost Converter, and Flyback Converter. By capturing fault progression quantitatively, the FSI enables a proactive approach to fault detection, facilitating targeted maintenance and component replacement in mission- and safety–critical applications.</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

Assessing fault severity in power converters with multiple component degradation using random forest regressor and statistical feature analysis

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Akanksha Chaturvedi,
  • Monalisa Sarma,
  • Sanjay K. Chaturvedi

摘要

In safety–critical applications, preventing the complete failure of power converters is crucial. However, existing anomaly detection methods primarily focus on single-component faults, neglecting practical scenarios where multiple components degrade simultaneously due to various stressors. To address this limitation, this paper presents a systematic methodology for detecting and analyzing anomaly states resulting from multi-component degradation in power converters. Statistical parameters derived from the output voltage are used to identify anomalies, capturing the system's dynamic changes as faults progress from single to multiple components. A random forest regressor is employed to determine the most significant features contributing to fault detection. Additionally, a Fault Severity Index (FSI) is computed to provide a quantitative measure of degradation severity. The methodology is validated through LTspice simulations on three converter types: Single-Ended Primary Inductor Converter (SEPIC), Boost Converter, and Flyback Converter. By capturing fault progression quantitatively, the FSI enables a proactive approach to fault detection, facilitating targeted maintenance and component replacement in mission- and safety–critical applications.