Label-free Industrial Fault Detection via Adversarial Inverse Reinforcement Learning: A System for Run-to-Failure Prognostics

2026-07-24

AI Summary

This work addresses the challenge of industrial fault detection under extreme label scarcity, where existing reinforcement learning approaches often neglect state transitions and temporal discounting, effectively degenerating into static classification. To overcome this limitation, the study introduces adversarial inverse reinforcement learning (AIRL) for unsupervised fault detection—a first in this domain—leveraging offline inverse reinforcement learning to directly recover an intrinsic “health” reward from observed state transitions, without requiring handcrafted rewards or fault labels. By doing so, the method transcends the constraints of conventional reconstruction-based models and contextual bandits, which struggle with dynamic system modeling and progressive degradation tracking. Evaluated on three run-to-failure datasets—HUMS2023, IMS, and XJTU-SY—the approach consistently achieves non-saturating post-detection alignment, significantly outperforming current state-of-the-art baselines.

Abstract

Machinery fault detection (MFD) remains heavily reliant on supervised learning, which struggles with the scarcity of fault labels in real-world settings. While reinforcement learning (RL) offers a framework to model the sequential nature of degradation, current ``RL-based'' MFD methods reduce the problem to a static contextual bandit (CB) formulation: by ignoring state transitions and discarding the temporal discount factor, they collapse to standard supervised classification. We propose an adversarial inverse reinforcement learning (AIRL) framework that treats MFD as an offline IRL problem. Unlike reconstruction-based approaches that rely on static error margins, or CBs that ignore dynamics, our method recovers an intrinsic "health" reward directly from observational state transitions, requiring neither manual reward engineering nor fault labels. On three run-to-failure benchmarks (HUMS2023, IMS, XJTU-SY), AIRL is the only method achieving non-saturated post-detection consistency across all datasets, while CB baselines fail to detect gradual degradation and reconstruction models collapse into always-anomalous states. Code and data: https://github.com/dhirajneupane/AIRL-MFD-DN.

Problem

• label-free
• industrial fault detection
• run-to-failure prognostics
• sequential degradation
• reward learning

Innovation

• Adversarial Inverse Reinforcement Learning
• Label-free Fault Detection
• Run-to-Failure Prognostics
• Offline IRL
• Machinery Health Reward