Influence of ea control parameters to optimization process of fjssp problem

Abstract

The ability of Evolution Algorithms (EA) to find an optimal solution is usually given by various algorithm operators. Population size and a maximal number of generations are usually set base on available timespan. Setting selection and elimination methods together with crossover probability are usually based on intuition and sometimes are problem specific. That is the reason presented research is focusing on the approach of how to set elimination methods and crossover probability by the statistical approach to eliminate the necessity of experience and intuition. This article describes the scheduling model together with the used EA to solve the Flexible Job Shop Scheduling Problem (FJSSP). Statistical process control methods are applied as there is a designed experiment to find out the statistical significance of each parameter during solving one of the FJSSP hardest problems. Crossover and elimination statistical importance are analysed and suitable levels of them are suggested. The statistical approach as a possible methodology to set the mentioned parameters is then discussed.