Scholarlink

<p>Hybrid power units (HPUs) operate under highly time-varying conditions and involve tightly coupled electromechanical subsystems, which makes accurate and timely fault diagnosis particularly challenging, especially when sensor faults coexist with power-performance degradation. To address this issue, this paper proposes a power-performance fault diagnosis framework for HPUs based on redundant monitoring and multi-source data fusion. A dual extended Kalman filter (DEKF) is developed to jointly estimate system states and sensor biases/parameters, generating physically consistent residuals under transient conditions. These residuals are then fused with raw sensor measurements through a fuzzy inference module to construct a robust diagnostic index. A logic- and threshold-based decision layer, further assisted by on-board diagnostic (OBD) codes, enables reliable fault detection and isolation. Simulation results demonstrate that, compared with a single-EKF baseline, the proposed method improves detection accuracy to 91.8%, reduces the false alarm rate to 2.8% and the missed detection rate to 3.2%, while shortening the average detection delay to 85 ms. Reliable fault isolation is maintained under simultaneous sensor and actuator faults, confirming the robustness and practical applicability of the proposed framework for hybrid power unit diagnostics.</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

Fault Diagnosis of Hybrid Power Unit Based on Redundant Monitoring and Data Fusion

  • 期刊论文

  • 发表日期：

    2026年3月31日

查看全文
  • Xiao Yuan Zhang,
  • Bo Yan Chen,
  • Hui Jing,
  • Gang Wang,
  • Cong Li,
  • Huan Qin Feng,
  • Zi Qiang Luo

摘要

Hybrid power units (HPUs) operate under highly time-varying conditions and involve tightly coupled electromechanical subsystems, which makes accurate and timely fault diagnosis particularly challenging, especially when sensor faults coexist with power-performance degradation. To address this issue, this paper proposes a power-performance fault diagnosis framework for HPUs based on redundant monitoring and multi-source data fusion. A dual extended Kalman filter (DEKF) is developed to jointly estimate system states and sensor biases/parameters, generating physically consistent residuals under transient conditions. These residuals are then fused with raw sensor measurements through a fuzzy inference module to construct a robust diagnostic index. A logic- and threshold-based decision layer, further assisted by on-board diagnostic (OBD) codes, enables reliable fault detection and isolation. Simulation results demonstrate that, compared with a single-EKF baseline, the proposed method improves detection accuracy to 91.8%, reduces the false alarm rate to 2.8% and the missed detection rate to 3.2%, while shortening the average detection delay to 85 ms. Reliable fault isolation is maintained under simultaneous sensor and actuator faults, confirming the robustness and practical applicability of the proposed framework for hybrid power unit diagnostics.