Section: New Results

Modeling and optimization of a launcher integration process

The following result has been obtained by Christophe Nivot (CQFD member), Benoîte De Saporta, François Dufour (CQFD member), Jacques Béhar, Damien Bérard-Bergery and Charles Elegbede.

We deal with the modeling and the optimization of a launcher integration process. The subassemblies go through various types of operations which are split up into workshops. Their operating time is supposed random due to possible breakdowns or staff issues. Storage capacity of output products is limited and costly. Launches have to be performed according to a predetermined schedule, and lateness also costs money. The rate of production of the subassemblies must be decided every year. Therefore, the system can be modeled with a Markov decision process which is suitable for decision optimization and cost minimization. Indeed, one must find a balance between slow production (thus low storage levels and high probability to be late), and fast production (high storage levels but respected schedule).

We propose a model of this integration process based on Markov decision models. We present the simulation we have performed so far and discuss the difficulties of the optimization.