Stochastic Scheduling for Batch Processes With Downstream Queue Time Constraints

This research studies the problems of stochastic dynamic scheduling in production systems with batch processes and process queue time (PQT) constraints. The production systems consist of upstream batch processing machines and downstream single processing machines. Under the PQT constraint, waiting time in the downstream queue is constrained by an upper limit and violating this constraint causes scraps of jobs. The batch process increases the probability of PQT constraint violation because a batch of work-in-processes (WIPs) will move simultaneously into the downstream queue after the service completion of batch processes and suffer from higher waiting time variance. A batch process admission control (BPAC) model is developed using Markov decision processes to minimize the sum of long-run average waiting and scrap costs. The proposed BPAC model explicitly considers uncertain factors in production systems given that uncertainties are major reasons for PQT constraint violation. These uncertain factors include job arrival, processing time, and machine breakdown/repair. To cope with these uncertain factors, the BPAC control decisions change dynamically with the real-time machine health and WIP distribution. The performance of BPAC is validated using discrete event simulation, and the simulation results confirm the significant performance improvement in a wide range of batch production environments.