Trustworthy and Explainable Deep Reinforcement Learning for Safe and Energy-Efficient Process Control: A Use Case in Industrial Compressed Air Systems

2025-12-20

AI Summary

Addressing the challenge of simultaneously ensuring safety, energy efficiency, and algorithmic trustworthiness in industrial compressed air system control, this paper proposes a trustworthy and interpretable deep reinforcement learning (DRL) controller. Methodologically, we design a multi-level interpretability pipeline integrating input perturbation robustness testing, gradient sensitivity analysis, and SHAP-based feature attribution to guarantee physically plausible decisions, prediction-driven control, and strict adherence to safety constraints. Our key contribution is the first deep integration of interpretability mechanisms into a closed-loop DRL industrial controller—without requiring explicit physical modeling. Experimental results demonstrate that, compared to conventional industrial controllers, the proposed approach significantly reduces overpressure events and achieves approximately 4% reduction in overall energy consumption. Attribution analysis further confirms that pressure and load forecasts serve as the primary decision drivers.

Technology Category

• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability
• Intelligent Robots: Behavior Learning & Control
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

This paper presents a trustworthy reinforcement learning approach for the control of industrial compressed air systems. We develop a framework that enables safe and energy-efficient operation under realistic boundary conditions and introduce a multi-level explainability pipeline combining input perturbation tests, gradient-based sensitivity analysis, and SHAP (SHapley Additive exPlanations) feature attribution. An empirical evaluation across multiple compressor configurations shows that the learned policy is physically plausible, anticipates future demand, and consistently respects system boundaries. Compared to the installed industrial controller, the proposed approach reduces unnecessary overpressure and achieves energy savings of approximately 4,% without relying on explicit physics models. The results further indicate that system pressure and forecast information dominate policy decisions, while compressor-level inputs play a secondary role. Overall, the combination of efficiency gains, predictive behavior, and transparent validation supports the trustworthy deployment of reinforcement learning in industrial energy systems.

Problem

• Develop trustworthy reinforcement learning for industrial compressed air control
• Ensure safe and energy-efficient operation under realistic conditions
• Provide explainable policy decisions through multi-level analysis methods

Innovation

• Trustworthy reinforcement learning for industrial compressed air control
• Multi-level explainability pipeline with SHAP and sensitivity analysis
• Energy savings without explicit physics models