Scheduling of multiproduct and multipurpose batch plants: Modeling and algorithmic developments

During the last two decades, the problem of production scheduling has emerged as one of the most significant challenges in the field of industrial plant operations, especially when multipurpose and multiproduct batch processes are involved and when there is the potential for significant savings through the use of mathematical modeling techniques. This thesis presents several fundamental contributions to the area of batch process scheduling including both mathematical models and algorithmic techniques used to solve difficult scheduling problems. The key problems addressed in the thesis are as follows.; I. Enhanced continuous-time formulation. The state-of-the art in unit-specific event based models for short-term scheduling is reviewed and comparisons are made with competitive global event based scheduling formulations. An enhanced continuous-time formulation is developed for the short-term scheduling of multipurpose batch plants with resource constraints, various storage policies, and intermediate due dates.; II. Closing the integrality gap. A new algorithmic technique is proposed to help reduce or even close the integrality gap experienced in large-scale scheduling problems. The proposed formulation employs a number of preprocessing steps and utilizes several sets of rigorous constraints to generate models with tighter LP relaxations. In addition, a new rigorous mathematical model is developed which can be used along with the proposed algorithmic techniques in order to solve large and difficult short-term scheduling problems.; III. Scheduling under uncertainty. A novel robust optimization methodology is proposed to address the problem of scheduling under uncertainty. When applied to Mixed-Integer Linear Programming (MILP) problems, the robust optimization formulation produces "robust" solutions which are, in a sense, immune against uncertainty in the inequality constraints. Robust optimization techniques are developed for several types of uncertain data including bounded uncertainty, bounded and symmetric uncertainty, and uncertainty described by a known distribution.; IV. Medium-term production scheduling. A mathematical model is developed for the problem of medium-term scheduling in a multipurpose, multiproduct industrial batch plant. The framework decomposes the large scheduling period into successive, short horizons of a few days and a novel continuous-time formulation for short-term scheduling of batch processes with multiple intermediate due dates is applied to each short horizon selected, leading to a large-scale, MILP problem.; V. Reactive scheduling. A mathematical framework is presented for the problem of reactive scheduling in the event of an unexpected unit shutdown or the addition or modification of orders. The formulation utilizes the short-term scheduling model developed for the industrial batch plant. To avoid full rescheduling, modifications are added to the scheduling formulation and a subset of tasks from the nominal schedule is fixed, allowing the full scheduling horizon to be updated in a reasonable amount of CPU time.; For each of the thesis topics, several example problems are presented to illustrate the effectiveness of the proposed formulation, and when appropriate, comparisons with other approaches from the literature are provided.