Research On Models And Algorithms For Multiproduct Joint Replenishment And Delivery Scheduling Problem

Joint replenishment is one of the most important procurement strategies,which can balance ordering cost and inventory cost to achieve better procurement performance.The procurement activities inevitably involve distribution planning.It is a better choice to coordinate procurement and distribution optimization rather than consider them separately.This dissertation focuses on joint replenishment strategy,coordinatedly considers delivery activity,and extends it to multi direction.On the basis of studying the basic joint replenishment and delivery(JRD)model,a better algorithm is designed.The basic JRD model is extended to three more general models:JRD model with quantity discount and resource constraints(CD-JRD),JRD with multi-warehouse(M-JRD),and coordinated dynamic demand lot-size and delivery scheduling problem(CDLSDP).Moreover,effective algorithms are designed for the new mathematical models.Firstly,in view of the deficiency of the method of solving basic JRD,an excellent lower bounds algorithm is designed for the first time by in-depth analysing mathematical properties of the model.Then,a variable neighborhood search(VNS)with the bounding method is developed to solve the basic JRD.A large number of computational examples,whether large-scale or small-scale,show that the bounding procedure can effectively and efficiently determine satisfactory bounds,and the VNS performs better than the best known heuristic and metaheuristic for the basic JRD.Secondly,considering the importance of quantity discount and resource constraints,the JRD model with quantity discount and resource constraints is formulated.Two algorithms are designed by in-depth analysing mathematical properties of the model:a modified RAND algorithm and a tabu search(TS)with the bounding method.A large number of computational examples,whether large-scale or small-scale,show that two algorithms can find satisfactory results.Meantime,we can find from the computational examples that when resource constraints is relaxed or the model is small-scale,quantity discount can lead to cost savings;while for large-scale model,limitation of resource constraints is obvious,the effect of quantity discount is weakened.Moreover,we find reaching the upper bound of resource constraints is easy because of the result of large procurement in the delivery.This indicates that imporving the resource constraints in the delivery can save cost more effective.Thirdly,a JRD model with multi-warehouse is designed by considering the practical procurement activities with multi-warehouse.Meantime,two algorithms are designed,i.e.a tabu search with RAND(TS-RAND)algorithm and an adaptive hybrid different evolution(AHDE)algorithm.A large number of computational examples show that TS-RAND algorithm can effectively and efficiently determine satisfactory results,while DE can only deal with small-scale examples.In addition,the results show that the total cost of M-JRD is significantly lower than the basic JRD.Lastly,a coordinated dynamic demand lot-size and delivery scheduling model is proposed in view of the continuous change of the demand.Then,a heuristic named four-phase(FP)is designed to solve the proposed CDLSDP by analysing mathematical properties of this model.A large number of computational examples show that FP algorithm can find excellent results and its efficiency is much higher than the classical mathematical programming method.CDLSDP provides a good implementation plan for the enterprises which have multi cycle purchasing activities.