Industrial Internet of Things-Based Prognostic Health Management: A Mean-Field Stochastic Game Approach

Abstract

Recent advances in industrial Internet of Things (IIoT) have dramatically leveraged prognostic health management for industrial systems. Indeed, the cognitive and communication capabilities of IIoT empower their integration in the industrial systems maintenance workflow to ease the transition toward industry 4.0. In this paper, we study a mean field stochastic game for IIoT-based CBM of industrial facilities formulated to favor grouped maintenance for cost reduction. We provide an analytical analysis of the proposed game to characterize its equilibrium operating point: mean-field equilibrium (MFE). We design a learning algorithm to reach the MFE based on a local adjustment of the maintenance rate and the global health state distribution of the monitored components. Numerical evaluation validates the proposed game and ensures maintaining a high fraction of the components in a healthy state by acting on preventive and corrective replacement rates.

Keywords

• Prognostic health management
• industrial Internet of Things
• mean-field stochastic games
• mean-field equilibrium
• H-learning
• Markov chain