Semiconductor manufacturing systems are known to be the most complex manufacturing systems in the world. In particular, the small size and complexity of the manufactured components impose time constraints (TCs). A TC corresponds to a maximum duration not to be exceeded between two manufacturing steps, that are not necessarily consecutive. TCs can have complex configurations, by succeeding or overlapping each other (time constraint tunnels, TCTs). The management of TCs, and in particular TCTs, is little covered in the literature on problems at the operational level, and non-existent at the tactical level to the best of our knowledge.

At the tactical level, a column generation approach using the notion of timed routes has been extended in this thesis in order to integrate TCs into the optimization of production plans. The modeling of timed routes at machine level, and the modeling of work-in-process were used to model and solve industrial instances. The impact of TCs on tactical plans was quantified, showing the relevance of integrating these constraints at this decision level.

At the operational level, a simulation model proposed in previous work has been significantly improved. Machines have been modeled more accurately, to take into account machines with serial and parallel chambers, and batch machines. Criticality indicators associated with TCs were also proposed, to complete the information provided by other performance indicators extracted from the simulation. The various indicators were used to propose TC and TCT management rules based on a maximum number of products in a TC group. These rules are created and improved using an iterative heuristic based on the indicators from the previous simulation, and are evaluated in the following simulation. The proposed optimization-simulation approach has been industrialized into a decision support tool to enable time constraint managers to validate and use the proposed approaches, and to automate the management of TCs and TCTs in the future.

Numerous perspectives are proposed both at the tactical level and at the operational level. In particular, various research avenues to reduce computational times and industrialize the column generation approach are presented.

Jury:

• YALAOUI Alice, Professor, Université de technologie de Troyes  (Reviewer)
• SAUER Nathalie, Professor, Université de Lorraine (Reviewer)
• ABSI Nabil, Professor, Mines Saint-Etienne (Examiner)
• KEDAD-SIDHOUM Safia, Professor, Conservatoire National des Arts et Métiers (Examiner)
• DAUZERE-PERES Stéphane, Professor, Mines Saint-Etienne (Thesis director)
• CHRIST Quentin, Doctor, STMicroelectronics (Industrial supervisor)
• ROUSSEL Renaud, Engineer, STMicroelectronics (Co-Industrial supervisor)
• VIALLETELLE Philippe, Engineer, STMicroelectronics (Guest)
• MOUSTAPHA SALEH Brahim, Engineer, STMicroelectronics (Guest.