Improving batch manufacturing process operations using mathematical programming based models

We describe the development and implementation of a hierarchy of process models used to improve operations in a batch manufacturing facility. Specifically, models have been developed to facilitate long-range planning and shop floor scheduling in a dedicated multiple stage batch processing facility. The objective of this work was to develop a systematic means of setting target production rates at each stage of the process to balance production resources and while meeting product demands and allowing the plant to be shut down for routine maintenance. A high level planning model was first developed to set target production rates at each stage of the process. Detailed operations taking place within each stage were not included in the high level planning model. Rather, each stage of the operation was considered as a task or small group of tasks with an average production rate and resource availability. This high level approach reduced the computational burden required to determine planned production rates over a long time horizon (3 months - 3 years). Since the planning model summarized the details of the relationships among the operations within a stage, it periodically set short-term target production rates for a particular stage that were quite aggressive. For this reason, detailed models were developed for the bottleneck stages within the operation. This allowed the bottleneck stages to achieve higher throughput so that production rates determined by the planning model could be met. Since the modeling and solution frameworks for the planning and scheduling models are common, the transition from the planning model to the scheduling model required a minimal amount of work. By combining the high level planning model with the detailed models for the bottleneck stages, we have developed a means of smoothing resource usage through the plant. Better planning for unusual conditions such as plant shutdowns and scheduling complex stages of the operation for the targets set by the planning model have been achieved. Furthermore, since a mathematical programming framework has been used, the feasibility of meeting a set of proposed demands can be directly verified.