Efficient Production Planning and Scheduling

During the 1980’s, a disturbing phenomenon was observed in both the United States as well as in Europe. Manufacturing jobs, which were once the staple of industrial economies, disappeared in alarming numbers. This coincided with a worldwide recession and massive unemployment.

Evolutionary algorithms refer to a class of adaptive methods which are based on evolution. In nature, populations of plants and animals evolve over many generations according to the principles of natural selection and the “survival of the fittest” (Darwin 1859). Evolutionary algorithms attempt to mimic this process of evolution in order to solve optimization problems.

Genetic algorithms have been successfully applied in engineering applications, such as the design of gas pipelines (Goldberg 1983) and telephone networks (Davis et al., 1993). They have also been successfully applied in the natural sciences, as illustrated by Unger and Moults work applying genetic algorithms to solve the protein folding problem (Unger and Moult 1993). Applications in business, however, have been relatively few (Nissen 1995) and limited to small prototypes or standard benchmark problems (Beasley 1990). Although benchmark problems provide a simplified means of comparing different optimization methods with one another, they are not very representative of industrial applications.

A feasibility study was conducted to investigate whether the hybrid genetic algorithm described in Chapter 4 could be successfully embedded within the AMTOS simulation environment.

A majority of the implementations of genetic algorithms described in the literature are for standard benchmark problems, such as the GA-testbed from Beasley (Beasley 1990). While benchmark problems provide a convenient means of comparison between different optimization methods, they are unfortunately not very representative of real-world problems.

This dissertation has examined the suitability of using a genetic algorithm in combination with simulation to improve production planning in an actual manufacturing environment. Although classical genetic algorithms have performed well in benchmark tests, there have been very few successful implementations of genetic algorithms in business (Nissen 1995). The challenge in this study was to find practical methods to adapt the theoretical constructs of classical genetic algorithms for use in a commercial application.