open access

https://hdl.handle.net/2115/100420

Stochastic MRAS-Based Strategy Optimization: Integration into PyCATSHOO, a Platform for Probabilistic Performance Assessment of Complex Systems

英語

PyCATSHOO

2026-07-01

Proceedings of the 10th International Symposium on Reliability Engineering and Risk Management (ISRERM 2026)

This paper investigates the integration of the stochastic Model Reference Adaptive Search (SMRAS) algorithm with a modeling framework based on piecewise deterministic Markov processes (PDMPs) for the analysis and optimization of complex hybrid systems. The proposed approach is implemented within the PyCATSHOO platform, which enables the modeling and simulation of systems combining discrete stochastic events with continuous deterministic dynamics. A case study is used to evaluate the proposed framework through the optimization of the maintenance policy of a multi-component system subject to aging effects. In this model, component failure rates depend on both the physical age and an apparent age representing cumulative operating time. Maintenance actions, such as replacement and refurbishment, may partially or completely reset these ages and thus reduce failure rates. The optimization problem accounts for several economic factors, including the costs of preventive maintenance actions, the costs of corrective maintenance following failures, and the revenues generated during system operation. These ele-ments are combined into a global performance criterion that must be maximized over the system’s operational lifetime. The SMRAS algorithm is used to explore the space of possible maintenance policies and to identify near-optimal policies. The results demonstrate the relevance of combining SMRAS with PDMP-based simulation models for maintenance optimization. They also highlight the sensitivity of the algorithm to the choice of certain hyperparameters, emphasizing the importance of appropriate calibration to ensure stable and reliable optimization results.

会議発表論文

Version of Record