Laboratoire de Génie Informatique et d’Automatique de l’Artois

At the heart of industrial efficiency lies the complex yet indispensable process of job-shop scheduling optimization. The focus of this doctoral thesis is to delve into this subject, particularly emphasizing the importance of preventive maintenance as a mechanism to minimize machine breakdowns and promote smooth Just-In-Time (JIT) efficiency.

The exploration of the intersection between classical optimization methods and Artificial Intelligence (AI), specifically reinforcement learning (RL), forms a central theme of the research. The objective is not only to harness the potential of these cutting-edge AI technologies but also to compare their efficacy against traditional methods in solving complex optimization problems.

The research journey begins with a focus on a single machine system, understanding the interplay between scheduling resumable jobs and preventive maintenance, and the impact of such considerations on reducing machine breakdowns and promoting JIT efficiency. Two algorithms were investigated. The first is based on a stochastic dynamic programming and the second is based on RL.

Moving beyond a single machine, the thesis expands to consider more complex settings like job-shop scheduling, taking into account the unique challenges of JIT delivery and operation-specific due dates. In this context, methodologies like Adaptive Large Neighborhood Search (ALNS) are studied, offering promising approaches for these NP-Hard problems.

Moreover, the proposed research proposes a comprehensive study on RL-based methodologies, demonstrating their robustness and effectiveness in optimizing job-shop scheduling while considering preventive maintenance in a JIT context. These findings contribute significant advancements to the field, providing a roadmap for future research focused on enhancing the performance and competitiveness of industrial production systems.