 |
International Society of Science and Applied Technologies |
| ACSA: A Hybrid Approach for Designing Smart Building Maintenance Strategies | ||||
| Author | Anas Hossini
|
|||
| Co-Author(s) | Leïla Kloul; Maël Guiraud; Benjamin Cohen Boulakia
|
|||
| Abstract | Effective Predictive Maintenance is essential for ensuring the reliability of Smart Building systems while minimizing maintenance costs. This paper explores maintenance optimization by modeling system interactions using a Fault Tree and characterizing component failures with Weibull distributions. We evaluate two optimization techniques to enhance decision-making in this complex environment: Reinforcement Learning and Simulated Annealing. We propose ACSA, a hybrid algorithm that integrates reinforcement learning (Actor-Critic) with stochastic optimization (Simulated Annealing) to adjust maintenance strategies dynamically. Experimental results show that ACSA achieves a superior balance between system reliability and intervention costs while meeting Quality of Service constraints. This hybrid approach leverages adaptive decision-making to efficiently manage maintenance in non-stationary environments, making it a scalable solution for interconnected systems or any system of systems.
|
|||
| Keywords | Smart Buildings, Predictive Maintenance, Reinforcement Learning, Simulated Annealing | |||
| Article #: RQD2025-271 | ||||
Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design |