Optimization Models And Algorithms For Supply Chain Under Disruptions

Supply chain is an integrated network that combines suppliers,distributors,retailers and end customers.From the view of minimizing cost in the long run,all partners should coordinate to make the strategic location decision,tactical ordering decision and operational scheduling decision.To reduce the overall operating costs and gain the competitive advantage,global supply chain network has enjoyed adequate development.However,the geographical dispersion and the structural complexity simultaneously bring vulnerability to the network.The global supply chain is easy to suffer from disruption risks,which makes the vulnerability an increasingly serious problem.Based on the existing research on classical supply chain network optimization problems,this dissertation first considers a more complicated but also realistic problem,by integrating location,inventory and routing decisions.A mixed integer programming model is constructed for the problem,a metaheuristic including initialization,intensification and post-optimization is designed.The performance of the algorithm is assessed through comparisons with published methods on the benchmark problems.Then,based on the aforementioned deterministic problem,this dissertation further incorporates facility disruption risks.Built facilities are subject to probabilistic disruptions.Once a facility fails,its customers should seek service from survived ones.Those reassignments bring extra emergency transportation costs.This dissertation considers disruption risks in the modeling process,so as to produce a more “reliable” network design plan — substantial improvements in reliability is attained with minimal increases in operational cost.This reliable location model is expanded from the following three aspects.First,the vehicle routing is introduced.Due to facility disruptions,customer reassignments will influence the present distribution routes,which will induce higher costs.For this two-stage stochastic problem,a simulated annealing based metaheuristic is designed.The heuristic includes a maximum-likelihood sampling method,routereallocation improvement and two-stage neighborhood search.Computational tests show that a more reliable plan usually exists.Then,the inventory cost is introduced.Disruption risks lead to solutions that have more open facilities than classical models,while inventory costs tend to produce solutions with fewer open facilities.By considering both inventory and disruptions,this dissertation can seek the appropriate balance between risk-diversification and risk-pooling.A nonlinear mixed integer programming model is developed,an exact approach using special ordered sets of type two(SOS2)and a heuristic based on Lagrangian relaxation are proposed.Interesting managerial insights are drawn through parametric sensitivity analysis.Last,the service competition is introduced.Two service providers — a leader and a follower — sequentially locate a fixed number of facilities,competing to capture market share.Each customer seeks the nearest operational facility for service.The problem is modeled as a Stackelberg game,and the leader's decision problem is formulated as a binary bilevel linear optimization problem.A variable neighborhood decomposition search heuristic is developed,and then further extended to solve general bilevel linear programs.