Contribution à l'ordonnancement post-pronostic de plates-formes hétérogènes et distribuées : approches discrète et continue.
Nathalie Herr · HAL (Le Centre pour la Communication Scientifique Directe) · 2015
This thesis addresses the problem of maximizing the production horizon of a heterogeneousdistributed platform composed of parallel machines and which has to provide a global productionservice. Each machine is supposed to be able to provide several throughputs corresponding todifferent operating conditions. It is assumed that using a machine with degraded performancescompared to nominal ones allows to extend its useful life before maintenance. The study fallswithin the decisional step of PHM (Prognostics and Health Management), in which a prognosticsphase allows to determine remaining useful lives of machines. The optimization problem consistsin determining the set of machines to use at each time and a running profile for each of them so asto maximize the production horizon before maintenance. Machines running profiles are definedon the basis of two models. First one depicts the behavior of machines used with a discretenumber of performances. For this case, the problem complexity is first studied considering manyvariants of the optimization problem. Several optimal and sub-optimal resolution methods areproposed to deal with the scheduling problem. Several sub-optimal resolution methods are thenproposed for the second model, which applies to machines whose throughput rate can varycontinuously between two bounds. These research works allow to determine the time beforefailure of a system on the basis of its components remaining useful lives.