Solution techniques for industrial-scale scheduling problems associated with batch production facilities

Matthew Harrison Bassett, Purdue University

Abstract

This thesis is concerned with quickly obtaining solutions for industrially relevant, large scale scheduling problems. A number of time based decomposition approaches are presented along with their associated strengths and weaknesses. It is shown that the most promising of the approaches utilizes a reverse rolling window in conjunction with a disaggregation heuristic. In this method, only a small subsection of the horizon is dealt with at a time, thus reducing the combinatorial complexity of the problem. The disaggregation heuristic further reduces the complexity by removing unlikely occurrences from consideration. Resource and task-unit based decompositions are also discussed as possible approaches to reduce the problem to manageable proportions. A number of examples are presented to clarify the discussion. With many industries changing to a just-in-time manufacturing environment, the ability to reliably meet due dates and/or determine delivery/promise dates is gaining importance. In order to achieve this, the ability to obtain realistic production plans is paramount. For these plans to be reliable, uncertainty needs to be incorporated. These uncertainties include, but are not limited to, processing time uncertainties, equipment reliability/availability, process yields, demands, and manpower fluctuations. A framework for including uncertainty by means of Monte Carlo sampling is presented. This framework is not limited to a specific model, but results obtained show that a model which provides a sufficiently accurate representation is necessary and is best satisfied by a scheduling model. A number of stopping criteria are derived and the framework utilized to obtain operating policies for two industrially based examples. Using concepts developed within this thesis, an expansion of a hierarchical decomposition approach for batch plant design is investigated using an equipment aggregation technique. This approach determines valid lower bounds (feasible solutions) to the overall maximization problem and uses these to reduce the overall solution time as the algorithm progresses. It also allows for finer discretization in the design stage (without complicating the combinatorics of the problems solved), leading to smaller, easier scheduling problems later. Methods for reduction in the number of market scenarios and equipment sizes specified are also discussed as a means to reduce the complexity of the problem.

Degree

Ph.D.

Advisors

Reklaitis, Purdue University.

Subject Area

Chemical engineering|Industrial engineering

COinS