Commande optimale stochastique appliquée à la maintenance des systèmes de production

Abstract

This thesis examines problems relating to the optimal planning of production systems subject to non-conforming and rejected products. Three optimal models are developed: 1. The first is based on knowledge of rejection rates at different productive asset states; 2. The second is based on the real-time control of non-conforming products, and 3. The last is based on the recovery of non-conforming products that are reused, much the same as the main raw material. In the first phase of this thesis, the first optimal model developed allows us to take into account the actual life cycle of all production system assets by counting the different operational levels of the system; thus breaking the system down into several modes of operation allows the current quantity of rejected products to be used to establish an optimal model which takes into account a combination of the asset age and human error due to the operations. The reject rate combines this group of failures into a single factor; using it in the model leads to cost-effective optimal and realistic production and maintenance policies. For the second phase of this thesis, a real-time control of the amounts of rejects seen during the production system operation allow us to establish an optimal preventive and corrective maintenance and production system; in which the rejected quantities serve as feedback, unlike most optimal production and maintenance systems, where productive asset age serves as the primary feedback. This consideration allows the establishment of cost-effective optimal production and maintenance policies, as compared to those developed taking into account just the age. In the last phase of this research, while controlling non-conforming quantities (rejects), reintegrating the rejects into the production system allows the establishment of an inverse optimal production and logistics system making it possible to significantly bring down demand for rare and expensive raw materials, apply opportunistic maintenance, carry out preventive maintenance, even where there is a disruption in inventory, in order to ensure that assets remain available and costs are reduced. The models developed in this thesis allow not only the control of asset age, but also their improper use, which invariably leads to human error. They are easily adaptable to industries with a precarious labor force, and where the turnover rate is very high, and leads to reduced assets availability.

Date	16 Dec 2011
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Jean-Pierre Kenné (Supervisor) & Robert Pellerin (Co-supervisor)