|
Date : Sept. 25, 2023, 1:30 p.m. - Type : Thesis - Jérémy BERTHIER - Amphi 1 Campus Georges Charpak
Modeling and solving multi-objective optimization problems for scheduling complex production systems |
The semiconductor sector has experienced uninterrupted growth since its emergence. Its industry has some of the most complex manufacturing processes, a complexity that is increased by the high degree of diversification of its products. Scheduling operations on the manufacturing floor is therefore particularly difficult given the wide range of constraints and the many operational objectives involved. In order to remedy this, optimization algorithms and mathematical programming are increasingly used for real-time scheduling of production operations, starting with the most critical workshops such as the photolithography workshop of the 300mm plant of STMicroelectronics Crolles considered for this work.
This thesis proposes contributions in terms of modelling and solving multi-objective optimization problems for scheduling. Two methods for accelerating the resolution of integer linear programming models for scheduling on parallel machines are introduced and validated using industrial data. Then, a formal framework for objective function comparison is defined and applied mathematically and experimentally to different scheduling criteria considered in the semiconductor industry. An analysis and redesign of the currently implemented objectives is also proposed. A generalization of the X-Factor function is introduced in order to integrate a contextual and tactical control of the scheduling solution used by STMicroelectronics. Finally, some other industrial and scientific contributions of the thesis are also presented.
This thesis proposes contributions in terms of modelling and solving multi-objective optimization problems for scheduling. Two methods for accelerating the resolution of integer linear programming models for scheduling on parallel machines are introduced and validated using industrial data. Then, a formal framework for objective function comparison is defined and applied mathematically and experimentally to different scheduling criteria considered in the semiconductor industry. An analysis and redesign of the currently implemented objectives is also proposed. A generalization of the X-Factor function is introduced in order to integrate a contextual and tactical control of the scheduling solution used by STMicroelectronics. Finally, some other industrial and scientific contributions of the thesis are also presented.
Keywords: operations research, multi-objective optimization, scheduling, mathematical programming, industrial engineering, semiconductor.
The thesis defense committee is composed of
- Safia Kedad-Sidhoum, Professor, CNAM (Reviewer) ;
- Vincent T'Kindt, Professor, Université de Tours ( Reviewer) ;
- Clarisse Dhaenens, Professor, Université de Lille (Examiner) ;
- Riad Aggoune, Doctor, LIST (Examiner) ;
- Stéphane Dauzère-Pérès, Professor, Mines Saint-Étienne (Thesis director) ;
- Claude Yugma, Professor, Mines Saint-Étienne (Co-thesis director) ;
- Antoine Perraudat, Doctor-engineer, STMicroelectronics (Industrial supervisor) ;
- Gaël Bergeret, IT manager, STMicroelectronics (Invited member) ;
- Rémi Poinas, IT manager, STMicroelectronics (Invited member).