Contribution to the dynamic rebalancing of assembly lines: modeling, resolutions and applications.

Summary If the balancing problem during the design of assembly lines is widely studied in the literature under the name of ALBP (Assembly Line Balancing Problems), few works concern their rebalancing in a short term and on-line horizon. The work presented in this thesis focuses on the dynamic rebalancing of assembly lines in response to short-term events disrupting production. The objective is to change the allocation of tasks on the workstations in order to smooth the load induced by the disruption. The short delay without production stoppage being more important than the optimality of the solution, the fast resolution time of the approximate resolution methods interested us. Two approaches were studied for their simplicity and novelty in the ALBP domain: the Iterated Local Search (ILS) method and the reachability search on a network of communicating automata. This showed us that they can meet the need for reactivity imposed by our problem but that they are difficult to adapt when the action levers to rebalance the line are modified. Our proposal to meet this need for adaptability is a methodology for modeling and solving the problems of rebalancing assembly lines based on a network of communicating automata and a search for attainability on it. To ensure genericity, the methodology is decomposed into two levels of abstraction: the study of the class of the ALBP problem considered and the resolution of the specific problem. The work has been validated on industrial examples from the Trane company.