A Bayesian network approach to job-shop rescheduling

Recently most researches in the scheduling area focus on scheduling under uncertainty, as real-world production activities are subject to considerable uncertainty. In highly dynamic situation, it is often required to resolve the baseline schedule. Deciding the right time to change the schedule becomes critical in minimizing the additional cost involved. The need for method that enables updating of current information and situation is absolutely required. In this paper we proposed a methodology to manage the shop- floor uncertainty using Bayesian network (BN). Although BN is widely used in several domains, the use of BN in manufacturing area is still uncommon. BN is a powerful approach for reasoning under uncertainty and it can be used to model the real time shop-floor condition. Here, we consider the schedule as a part of the total system. Hence, the proposed model considers both direct and indirect factors, i.e. it includes the interaction of the schedule with other factors in the system. Furthermore, BN is extended into influence diagram to evaluate the need of rescheduling. In addition, a different approach is proposed to define the conditional probability of the nodes that need further analysis. The proposed method is applied to the case of stochastic job-shop scheduling systems.