Enhancing Trustworthiness in AI-Based Prognostics: A Comprehensive Review of Explainable AI for PHM

Prognostics and Health Management (PHM) has emerged as an essential field for ensuring the reliability and safety of critical systems, increasingly becoming a key component in their maintenance and operation. While artificial intelligence (AI) has become a powerful tool for PHM, the opaque nature of AI methods might present hurdles in their deployment for real-world PHM applications, as the lack of transparency in how these methods process and interpret data. In this light, Explainable AI (XAI) is a rapidly growing field that aims to address those challenges. This understanding enhances the trustworthiness and acceptability of AI-based PHM solutions. Nevertheless, inconsistency in terminologies and concepts across diverse studies on XAI in PHM persists, accompanied by the absence of a standardized taxonomy. Our study aims to address this gap by presenting a comprehensive review of XAI in PHM, clarifying key concepts, and introducing a structured taxonomy tailored for PHM applications. This taxonomy facilitates the effective categorization of XAI methods, guiding their selection based on specific PHM requirements. We also delve into the delicate balance between the model’s explainability and performance, emphasizing XAI’s role in ensuring algorithmic fairness and reliability in PHM solutions. Additionally, the paper outlines practical challenges and potential future research directions in XAI for PHM. Overall, this contribution significantly advances the understanding and application of XAI in PHM practices.