Research On Dynamic Facility Location With Inventory Cost Of Distribution Centers

Global optimized logistics system is the basis of successful supply chain distribution management. The logistics decisions of distribution stage can be classified as distribution centers location decisions, distribution network's transportation decisions and distribution centers'inventory decisions by different functions in logistics system. Many influence factors change by time when we make decisions for this problems. Especially, for the distribution centers location decisions, key factors such as customer demands, transportation cost show dynamic characteristics clealy because of long planning time. Traditionally, many distribution centers location decisions models ignored dynamic characteristics of influence factors, because of the complexity of developing and sovling the mathematical models. This could result in suboptimal "myopia" decisions. It is necessary that the dynamic decisions environment must be considered in distribution centers location decisions models to avoid this problem.The three logistics decisions problems of distribution stage aforementioned have differences in contents, range and planning time, so these decisions are made in time sequence traditionally. For example, the distribution centers location decisions are made firstly, then the inventory decisions. However, these decisions have influence on each other's contents and there are many reverse benefits relationships between them. Therefore making decisions in time sequence don't ensure that logistics system is global optimal. To make logistics chain harmony and improve whole logitics system efficiency, it is needed to make decisions of part or all these problems simultaneously, such as location-inventory decisions, inventory-transportation decisions and location-inventory-route decisions and so on.In this thesis, dynamic distribution centers location problem was researched with considering the effect of inventory cost on location decision. The dynamic distribution centers location models with inventory cost was developed by using mathematics programming methods. The models included the reverse benefits relationships between location, transportation and inventory. After solving the models, the locations, numbers, assignment, order batchs and order frequency of distribution centers in ever period was obtained. These decision parameters are useful to build logistics system that is adaptable dynamic decision environment and has high global efficiency. The genetic algorithm, clone celection algorithm, particle swarm optimization algorithm and ant system were used to sovle the developed models and the capacities of finding optimal solution, counting speed and stability were compared between these algorithms to find the most suitable one. The concrete research results of the paper included:1. Aiming at the problems that the distribution centers original building cost was counted repetitionally(the models of distribution centers opened gradually) and the distribution centers original opening cost, closing cost, re-opening cost and operational cost were not completely considered(the models of distribution centers opened and closed more than once) in traditional dynamic distribution centers location models, the rational computing methods of distribution centers original building cost and complete computing methods of the distribution centers original opening cost, closing cost, re-opening cost and operational cost were developed.2. Aiming at the problem that dynamic decision environment was not considered in existed location-inventory models, the dynamic location-inventory models of distribution centers opened gradually and distribution centers opened and closed more than once were developed respectively corresponding distribution centers built by company and hired. The first model supposed that once distribution centers were built, they were not allowed to be closed and the results were the distribution centers opening schedulings. The second model supposed that once distribution centers were open, they coud be closed and re-opened. This coud better respond to dynamic decision environment. Four intelligent optimizing algorithms were used to solve the models. The results showed that the developed two models were feasible, and genetic algorithm was the most suitable algorithm when the problem had smaller size, and particle swarm optimization algorithm was the most suitable algorithm when the problem had larger size. After that, the inventory cost constrained by distribution centers storage capacity was analyzed and the dynamic location-inventory model with storage capacity constraints was developed.3. The dynamic location-inventory decision problems with multi-products were researched based on the dynamic location-inventory decision problems with single product. According to different multi-products inventory models, the dynamic location-inventory decision problems with multi-products under "independent cycle" hypothesis, "fixed cycle" hypothesis and "integer times cycle" hypothesis were developed respectively. Because the multi-products inventory models under "independent cycle" hypothesis and "fixed cycle" hypothesis were similar, the two models were combined into dynamic multi-products location-inventory models under "independent cycle and fixed cycle" hypothesis. Four intelligent optimizing algorithms were used to solve the models again. The results showed that the developed two models were feasible, and particle swarm optimization algorithm was the most suitable algorithm for dynamic multi-products location-inventory models under "independent cycle and fixed cycle" hypothesis, and for the dynamic multi-products location-inventory models under "integer times cycle" hypothesis, genetic algorithm was the most suitable algorithm when the problem had smaller sizes, and particle swarm optimization algorithm was the most suitable algorithm when the problem had larger sizes.4. Two decision methods were compared quantificationally. One method is making location decisions first and then making inventory descisions, the other is making loacaion and inventory decisions at the same time. It was proved that the latter was better than the former in total cost. Firstly, the dynamic facility location models without inventory cost and with inventory cost were developed. Secondly, numeric examples were simulated and analyzed and the results show latter method's total cost was less than former method's while increasing order cost or inventory holding cost per unit, or decreasing transportation cost per unit. And this savings of total cost were realized through decreasing total location cost and transportation cost.The research results of this thesis increased the contents of logistics system optimization theory, provided a rational decision method for decision makers to construct global optimal and adaptable logistics system, and so were valuable in practice.