Explainable Deep Reinforcement Learning Enhancing Industrial Maintenance
Year: 2025
Type: article
Abstract: Industrial maintenance is crucial for operational efficiency, requiring precise and timely decision-making to minimise downtime and costs while preserving system reliability. This paper presents a framework that integrates Deep Reinforcement Learning (DRL) with Explainable AI (XAI) to optimise preventive maintenance strategies and enhance interpretability for human operators. By applying XAI methods to DRL-driven decisions, the framework offers transparent insights into the rationale behind recommended actions, bridging the gap between AI-generated output and operator trust. Our work contributes to the post-hoc interpretability branch of Explainable Reinforcement Learning (XRL) and aligns with the objective of making DRL models both powerful and comprehensible in high-stakes industrial contexts. The simulation results demonstrate notable reductions in downtime, along with improved system reliability and operational transparency. These fndings underscore the potential of merging DRL and XAI in industrial maintenance, creating opportunities for more effective and trustworthy AI solutions in real-world settings.
Source: IFAC-PapersOnLine
Institution: University of Naples Federico II
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4414567717