Condition-Based Production for Stochastically Deteriorating Systems: Optimal Policies and Learning

Collin Drent,Melvin Drent,Joachim Arts

2023-12-06

Abstract:Production systems deteriorate stochastically due to usage and may eventually break down, resulting in high maintenance costs at scheduled maintenance moments. This deterioration behavior is affected by the system's production rate. While producing at a higher rate generates more revenue, the system may also deteriorate faster. Production should thus be controlled dynamically to trade-off deterioration and revenue accumulation in between maintenance moments. We study systems for which the relation between production and deterioration is known and the same for each system as well as systems for which this relation differs from system to system and needs to be learned on-the-fly. The decision problem is to find the optimal production policy given planned maintenance moments (operational) and the optimal interval length between such maintenance moments (tactical). For systems with a known production-deterioration relation, we cast the operational decision problem as a continuous-time Markov decision process and prove that the optimal policy has intuitive monotonic properties. We also present sufficient conditions for the optimality of bang-bang policies and we partially characterize the structure of the optimal interval length, thereby enabling efficient joint optimization of the operational and tactical decision problem. For systems that exhibit variability in their production-deterioration relations, we propose a Bayesian procedure to learn the unknown deterioration rate under any production policy. Our extensive numerical study indicates significant profit increases of our approaches compared to the state-of-the-art.

Optimization and Control,Probability

What problem does this paper attempt to address?

The paper attempts to address the problem of how to dynamically adjust productivity in a randomly degrading production system to balance revenue accumulation and degradation risk, and to optimize production decisions between planned maintenance times. Specifically, the paper studies the following two aspects:

1. **Known degradation relationship**:
- When the degradation relationship of the system is known and the same for each system, how to find the optimal production strategy (operational level) and the best maintenance interval length (tactical level).
- The authors model the decision problem at the operational level as a continuous-time Markov decision process (MDP) and prove that the optimal strategy has an intuitive monotonic property.
- Sufficient conditions are proposed to verify the optimality of the "bang-bang" strategy (i.e., setting productivity to either the maximum value or shutting down), and the structure of the optimal maintenance interval length is partially described.

2. **Unknown degradation relationship**:
- When the degradation relationship of the system varies by system and needs to be learned in real-time, how to find the optimal production strategy.
- The authors propose a Bayesian approach to learn the unknown degradation rate and construct a heuristic strategy.
- Numerical studies show that this heuristic strategy performs close to the Oracle strategy with known degradation relationships in the "bang-bang" region and is robust to variations between systems.

Through these studies, the paper aims to provide managers with a framework to dynamically adjust production decisions based on real-time state monitoring and integrate them with maintenance plans to maximize profits.