A study of heuristic scheduling

Abstract

This thesis is concerned with the subject of heuristic scheduling in the context of medium-sized batch and jobbing manufacture.
A scheduling program involving up to 1500 jobs and 150 machines is described and the methods by which:conflicts
were resolved and objective functions were measured is discussed. An account is given of the computation experience
found in applying the proposed scheduling technique to 15 different problems.
Precise despatching parameter control is applied to jobshop scheduling problems and the results presented in the form of reponse surfaces. The properties of the measured responses are examined using the techniques of Fourier and Walsh analysis and a model of 'regular' constituents of the surfaces is presented.
The re:sponzue surfaces observed in job-shop scheduling problems are compared with the theoretical respcnse of a simple f lowshop problem and the points of similarity discussed Classical non-linear optimization techniques are introduced
and applied to job-shop scheduling problems. A new approach to the optimization of schedule response surfaces is presented and compared with classical optimization techniques using problems of different sizes and complexity. -.
The proposed optimization technique was found to be unsuitable for small problems involving less than 10 jobs and 10
machines. For problems larger than this size, the technique was found to stabilize and the number of iterations was
reduced, typically, by a third when compared with the best classical optimization technique.