4 October 2025 in Green Energy and Intelligent Transportation
4 June 2026 in IEEE Sensors Journal
Equipment such as dispensers and compressors in hydrogen refueling stations operates under prolonged high-pressure and high-frequency conditions. Traditional fault diagnosis methods for hydrogen-related equipment typically rely on data-driven deep learning models.Although these models can achieve high diagnostic accuracy, their “black-box” nature hinders fault cause tracing. Moreover, they cannot verify the reliability of the inference process, which limits their interpretability. To address these issues, this study proposes an interpretable equipment-level fault diagnosis method based on sensor data. First, a fuzzy parameter optimization algorithm is developed to jointly optimize the antecedent and consequent parameters of a multi-rule-based Takagi–Sugeno–Kang (TSK) fuzzy system. The diagnostic accuracy reaches 96.54%. Second, an interpretability matrix is constructed to characterize the relationships between sensor data and fuzzy rules, facilitating fault cause tracing. In addition, an inference-based interpretability method is proposed to reveal the internal logical chain from multi-sensor inputs to the final diagnostic results, thereby verifying the accuracy of the inference process. Finally, a case study is conducted using real multi-sensor monitoring data collected from an operating hydrogen refueling station. The results demonstrate that the proposed method maintains high diagnostic accuracy, while providing clear interpretability in decision-making and inference processes. Thus, the proposed method can support the reliable operation and maintenance of hydrogen energy equipment.