A modelling of genetic algorithm for inventory routing problem simulation optimisation

• Main Authors: Othman, S. N., Mustaffa, Noorfa Haszlinna, Radzi, N.H.M., Sallehuddin, Roselina, Bazin, Nor Erne Nazira
• Format: Article
• Published: ExcelingTech 2016
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/71269/; https://www.scopus.com/inward/record.uri?eid=2-s2.0-85018409245&partnerID=40&md5=fa0fb832489d3e63913f429190f1f0eb
• Institution: Universiti Teknologi Malaysia

This paper presents the simulation optimization modelling for Inventory Routing Problem (IRP) using Genetic Algorithm method. The IRP simulation model is based on the stochastic periodic Can-Deliver policy that allows early replenishment for the retailers who have reached the can-deliver level and consolidates the delivery with other retailers that have reached or fallen below the must-deliver level. The Genetic Algorithm is integrated into the IRP simulation model as optimizer in effort to determine the optimal inventory control parameters that minimized the total cost. This study implemented a Taguchi Method for the experimental design to evaluate the GA performance for different combination of population and mutation rate and to determine the best parameters setting for GA with respect to the computational time and best generation number on determining the optimal inventory control. The result shows that the variations of the mutation rate parameter significantly affect the performance of IRP model compared to population size at 95%confidence level. The implementation of elite preservation during the mutation stage is able to improve the performance of GA by keeping the best solution and used for generating the next population. The results indicated that the best generation number is obtained at GA configuration settings on large population sizes (100) with low mutation rates(0.08). The study also affirms the premature convergence problem faced in GA that required improvement by integrating with the neighbourhood search approach.