DynamIc control and planning for intelLIGENT manufacturing processes and systems – DILIGENT
/
Part of the work addressed the flexible workshop scheduling problem by taking into account transport resources. Operations must be allocated and planned on machines and vehicles, and vehicle routes determined. To model this problem, an extension of the classical disjunctive graph model was proposed, including transport operations. A meta-heuristic was subsequently developed. Experimental results on new benchmarks demonstrate the effectiveness of this approach, the quality of the motion evaluation procedure, and the relevance of explicit modeling of transport resources. Another part of the work addressed the scheduling problem on parallel metrology machines to minimize risks. Several optimization methods were developed, such as a meta-heuristic (simulated annealing) and a column generation approach. An indicator named Global Sampling Indicator (GSI) was defined to dynamically model risks. Batches to be measured are selected to minimize the overall risk in the factory. Experimental results show that the column generation approach is efficient, but can be improved by dynamic programming for some cases, especially those with high variability in batch qualification.
/
Berterottière, L.; Dauzère-Pérès, S.; Yugma, C. Flexible job-shop scheduling with transportation resources. European Journal of Operational Research. 2024, 312, 3, 890-909.
Martin, M.; Dauzère-Pérès, S.; Yugma, C. Scheduling on parallel metrology tools for risk reduction in semiconductor manufacturing. ROADEF 2024-25ème congrès annuel de la Société Française de Recherche Opérationnelle et d'Aide à la Décision, Amiens, France. 4-7 Mars 2024.
Following the success of IMAGINE, the consortium aims to extend the integrated decisions from the real-time configuration to the predictive strategy. For intelligent manufacturing processes and systems, dynamic dispatching/scheduling and process/equipment control should be managed in a prognostic way. Production plans are dynamically optimized and used as the foundation for the predictive regulations of the process and equipment. As the production and engineering data are accumulated continuously, the predictive control and planning strategies will be empowered by the deep learning techniques that make sense of big data for accurate and precise previsions. Furthermore, the prognostic decisions will be interpreted with reasonable logic and traced back to the controllable factors for not only comprehending the analytic causality but also enhancing the control and planning dynamics. The methods developed in this project will be validated through the cooperation with local partners.
Project coordination
Claude YUGMA (Laboratoire d'Informatique, de Modélisation et d'Optimisation des Systèmes)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
Partnership
LIMOS Laboratoire d'Informatique, de Modélisation et d'Optimisation des Systèmes
National Taiwan University / Graduade Institute of Industrial Engineering
Help of the ANR 213,996 euros
Beginning and duration of the scientific project:
- 36 Months
