ABSTRACT
A common problem for scheduling systems in the production industry is the assignment of customer orders to shop floor processing (i.e. deciding the sequence of orders to be carried out). The majority of research in production scheduling has used time-based performance measures to evaluate various proposed scheduling strategies, such as lateness, tardiness, flow time, percent tardy, and due dates. However, in a practical factory the production performance can not often be measured in isolation, especially when there exists conflict between the measures. Therefore, it is desirable to have a single approach that can consider multiple performance criteria at the same time. Itoh (1993)
proposed a method to validate multiple performance criteria, which is based on heuristic dispatch. Sequence scheduling has long been seen as a problem that belongs to the class of NP-complete problem. DeJong and Spears (1989) use genetic algorithms (GA) to investigate such scheduling problems. In GAs, a objective function acts as a survival environment for their candidates, which are normally encoded as genes in a string, or chromosome. A random population pool of chromosomes is created as a first generation. GAs select pairs of individuals from this pool based on their performance in optimising the objective function. The population of low fitness chromosomes decreases over generations, while the population of high fitness ones increases and is selected to bear many offspring from the simulated evolution generation. This procedure is repeated in successive generations resulting in inferior traits in the pool die out due to lack of reproduction, while strong traits tend to combine with other strong traits to produce children.
