Forward Management of Spare Parts Stock Shortages Via Causal Reasoning Using Reinforcement Learning
Résumé
Our role in this paper was to search the appropriate means serving as a decision support tool for the choice of a policy and procurement planning of spare parts, contributing to the maintenance in operational conditions of industrial equipment and enabling to avoid stock outs and all at a lower cost. For this, we present a generic Bayesian model of consumption of spare parts in a replenishment policy type (T, s, S) adapted (mT, s *, S). The originality of this research is the fact that we characterize the process of consumption of spare parts by a set of typical scenarios, called "consumption configurations" and identified by a system of performance indicators using variables state in a Bayesian model. After defining all these indicators, the research enchain to deploy a Bayesian network which allow, through Bayesian simulation, obtaining a replenishment planning indicating the optimal combination by period: the durations of these replenishment periods, quantities to purchased, types of SP (new or revalorized), costs and associated risk of rupture, purchasing costs and induced storage.