Multi-Objective Optimization Of Closed-Loop Logistics Network With Facility Expansion

With the increasing exploitation of resource, environment is getting worse andworse, resource saving and environment protection have been a critical issue need to besolved nowadays. Due to the enacting of related policies and laws, more and moreenterprises are forced to dispose the waste products. Closed-loop logistics (CLL) givesthe companies an opportunity to solve this problem. In consequence, building a highefficiency and low cost CLL network becomes the strategic focus of these firms.CLL network is the integration of traditional forward logistics and recent reverselogistics, which includes factories, distribution centers, consumers, collecting centers,recycling centers, remanufacturing centers and so on. How to reasonably locate thefacilities and allocate the products in the network has been studied by more and morescholars as well as enterprises.This paper structures a CLL network with facility expansion based on multi-period,multi-product, and multi-stage. Differs with other CLL networks, the distributioncenters, recycling centers, and remanufacturing centers of this network can be expanded.As for the optimization objectives, network cost and network response are involved inthis model. Network cost includes fixed set-up cost, expansion cost, transportation costand processing cost, network response mainly consider the total delay time in deliveryand collection processes.The multi-objective optimization problem in the proposed model is a NP-hardproblem. With the increase scale of this NP-hard problem, it takes long time to solve it.As a result, a two-stage Evolutionary Algorithm is presented, and a Greedy Algorithm isadopted in the solving procedure to calculate the flows between the facilities. Unlike thesingle-objective problem which has only one optimal solution, the multi-optimizationproblem has a Pareto solution set. The Arena’s Principal method is used to update thetentative Pareto solution set through evolutionary iterations. Finally numericalsimulation is implemented to verify the convergence, effectiveness and efficiency of thealgorithm.