A Multi-Objective Model to Single-Allocation Ordered Hub Location Problems by Genetic Algorithm

Most of supply chain and Hub location problems involve several conflicting objectives hence requiring a multi-objective formulation. Normally, Multi-objective approaches lead to the maximization of a weighted sum of score functions. Since normalizing these functions and quantifying the weights is not a straightforward process, such approaches are poor in practice. In this research, this difficulty is overcome by using a modified genetic algorithm for evaluation of solutions. Several qualitative and quantitative objectives are considered referring to layout model that also allows practical constraints take into account. Due to these constraints, many of strings in the population resulting from this model may be in infeasible reigns; the common approach to solve such problems is to omit infusible solutions. However, as these solutions my have useful criteria that can improve the average fitness of the population, they can be used to achieve better solutions. The proposed model uses a graded penalty term to penalize infeasible solutions to pressure the search towards feasible regions and subsequently uses their useful criteria.