Reactive scheduling of production flow shop for accomodation of a new order

Abstract

As an archetype of the production systems, the flow-shops are featured by their highly dynamic behaviour caused by random disruptions such as machine breakdown, frequent arrivals of new order, tool breakdown, etc. In such cases, the companies need to react in an optimal way, thus the disruptions are discarded or managed and the economic criteria are maintained or even improved. In this context, reactive scheduling can justifiably be considered as a decision support tool that helps the operational managers to mitigate the adverse effects of disruptions occurring during the execution of initial schedule. Among these disruptions, the random arrival of new orders is the most frequently occurring phenomenon in the production shop floor. This random order arrival causes the non-optimality and/or infeasibility in schedule, which reduces resource utilization, increases work in progress (WIP) inventories and subsequently lowers the profit. Hence to aid the manufacturers, in this study a reactive flow-shop scheduling problem subject to the arrival of new order is addressed comprehensively, analysed critically and solved optimally. Though a number of researchers have already proposed the reactive scheduling models, most of them are complex and even difficult to execute within the acceptable time horizon. Therefore in this research, based on a widely adopted and renowned Wagner formulation, a Mixed Integer Non Linear reactive scheduling model is proposed. To ease the implementation, a predictive scheduling model is also developed together with the reactive one. The proposed models are implemented in flow shop of a local automotive industry and assessed by using four performance measures namely the makespan, mean idle time, mean waiting time and resource utilization. For both models the makespan is taken as the main objective function, while constraints such as capacity of the machines, completion time of the jobs and precedence relationships are formulated. The entire model is designed in a spreadsheet and solved by ‘What’s Best’ optimizer. The output obtained by comparing with the current practices demonstrates that the proposed approach has supremacy in reducing the makespan by 26.92% through predictive and 24.07% through the reactive scheduling. Additionally, a considerable reduction in the waiting time (72.22 % and 30 %) is attained through the implementation of the proposed approaches. Meanwhile, the utilization of the resources is also enhanced by 36.84 % through predictive and 61.9 % through reactive schedule thus yielding better productivity of the shop floor. It is expected that the implementation of the proposed approach would help improve the decision making procedure in automotive as well as other manufacturing industries.