A hybrid algorithm of tabu search and benders decomposition for multi-product production distribution network design

A multi-product two-stage production distribution network design problem (PDND) is considered in this paper. The network consists of capacitated plants, capacitated distribution centers and customers. The known demand for multiple products at the customers is satisfied by shipping via distribution centers, and each customer demand is supplied via a single distribution center. A relatively efficient hybrid algorithm of tabu search and Benders decomposition is proposed to determine where distribution centers are located and by which distribution center each customer is served exclusively. In Benders decomposition framework, a tabu search algorithm is used to solve the master problem to obtain quickly approximate-optimal solutions. At the same time, three kinds of neighbourhood structure are used in tabu search. The computational experiments show that the hybrid algorithm saves tremendous computation time in comparison with conventional Benders decomposition.