A Hybrid Evolutionary Algorithm And Its Application On Production Scheduling

Engineering design owns the characteristics of complexity and multi-constraints. Usually, problems are given as multi-objective which needs to be optimized on the feasible region. One of the most representative issues is production scheduling which is studied widely these days.In semiconductor manufacturing, multi-type of workspaces, huge and variable products which make the processing procedure extraordinary complicated, especially, it comes to a difficulty on rescheduling dynamic and effectively with reentrant mechanism. As an example of wafer making, we propose a hybrid algorithm based on GA (genetic algorithm) through changing the way of mutation which aims at reducing the total flow time. Moreover, the proposal is much effective on rescheduling when some sudden circumstances taken place.Multi-objective optimization is widely used on flow shop scheduling problem. As the particular of multi-objective, we propose a hybrid evolutionary algorithm based on local search method, mainly to find a balance between the computing time and solution deterioration. The most effective advantage for hybrid evolutionary algorithm is the improvement of convergence speed, but with the negative effect of the computing time for each generation is increased. Therefore, the global search is not fully utilized. As a result, it becomes much important for balancing the local search and genetic search. This paper addressed a local search based hybrid multi-objective evolutionary algorithm for flow shop scheduling problem. The fitness value of this proposal is decided both by its rank and the crowding distance between two adjacent chromosomes together. Firstly, we rank the order in the population and proposed a hybrid local search for decreasing the running time and improving the convergence speed to maintain the diversity of population. Then a binary tournament selection method is used for generating a new set of mating pool after local search operation. For generating new population, two-point crossover and selection mutation are introduced. Finally, some comparisons are made among the proposed approach with the other two mainstream algorithms for multi-objective problem. The experiment shows that the proposed method is much effective on the convergence speed.