Research Of Model And Algorithm Of Logistics System

Logistics is the third profit source of enterprises and it is the artery and basic industry of national economy. Application of information technology in logistics system can reduce logistics cost efficiently. Model and algorithm of logistics are hot research topics of logistics science.Logistics cost mainly consists of that of location, distribution and inventory. To reduce logistics cost, location, distribution and inventory of logistics system are studied in this dissertation.The work and innovation of this dissertation mainly consist of:1. Location problem of a single emergency center is studied. To locate an emergency center, not only should construction cost and transportation cost be considered, but the coverage and the time restriction from the center to the demand points must be taken into account. Gravity method, analytical hierarchy process (AHP) and solving location formulation are usually used to solve location problem of emergency center. But time from center to demand points and other influencing factors can't be evaluated in gravity method. AHP, alone, must evaluate all the candidate locations and other algorithm must be used to get the cost and time spent from each candidate location to the demand points, therefore, the work is too much. While location formulation is either too complex and hard to get solution by analytic method or unable to take all factors into account.To reduced location cost of emergency center, a three-stage algorithm (TSA) is proposed based on gravity method, AHP and the K-shortest path algorithm proposed in this dissertation. Location range is reduced by gravity method at the first stage. At the second stage, a K-shortest path algorithm, which is with little time complexity and easy to program, is proposed firstly and then it is used to eliminate those candidate points that can't satisfy the time restriction from the candidate points set. At the third stage, each remaining point in the candidate points set is analyzed by AHP, and a single point is chosen to be the final location of emergency center.Though TSA is simple and less amount in the calculation, many influencing factors like location cost, coverage, time restriction and other factors are taken into cosideration. So it is an efficient algorithm to solve location problem of a single emergence center.2. Three problems in logistics distribution are studied:(1) To minimize the traveling distance of vehicles, capacitated single-depot vehicle routing problem (SDVRP) based on coordinates of points is studied. Exact algorithm, heuristics algorithm and meta-heuristics algorithm are commonly used to solve this problem. Exact algorithm usually consists of large amount of calculation, and heuristics algorithm commonly can't get optimal solution. Meta-heuristics algorithm is widely used in many papers, among which genetic algorithm covers a large proportion. But the shortcoming of simple genetic algorithm that is easy to trapping into local optimum (prematurity) is not overcomed commendably in the existing research achieves.A hybrid genetic algorithm is designed to solve Capacitated SDVRP. A new crossover policy is proposed to overcome the prematurity of simple genetic algorithm. Better individuals preserved policy is adopted and a function is designed to decide the number of better individuals. To accelerate the procedure of the evolution, optimal policy is shemed out according to locations (?)lations between points and some individual are selected to optimize in each generation according to given probability.This algorithm efficiently overcomes the prematurity of simple genetic algorithm, and also has good performance.(2)Capacitated SDVRP based on transportation network is studied.Logistics distribution is usually performed in transportation network, and its cost can't be solely evaluated by traveling distance of vehicle. Few researchers have paid attention to this problem.A least distribution cost model of capacitated SDVRP based on transportation network is formulated, and a hybrid Parthenon-genetic algorithm is designed to solve the model. An optimization policy is proposed to accelerate the procedure of the evolution, which optimized individuals by inserting the shortest path based on transportation cost between two points.(3)Multi-depot vehicle routing problem(MDVRP) is studied. Few researchers have focused on MDVRP, and MDVRP is commonly converted to SDVRP to solve in the existing research achievements. Ant colony algorithm (ACA) has such merits as robustness and easiness to be integrated with other methods, and has been used in SDVRP successfully. But it hasn't been reported that ACA is used to solve MDVRP.A hybrid ant colony algorithm (HACA) is proposed to solve MDVRP successfully by ants' transfer policy and algorithm to construct solution designed in this dissertation. To improve HACA's performance, K-neighbor is used to restrict the ants' transfer objects, and 2-Opt is used to optimize ants' routes and solution, and regulation of pheromone update is designed. Hence, HACA succeeds in applying ACA to the solution of MDVRP, and also has a good performance.By data experiments these three algorithms' validity is verified, and the influence of parameters on algorithms' performance has been discussed.3. Stochastic inventory control problem of merchandise with shelf-life is sdudied. Analytic formulation and computer simulation are often used to solve inventory control problem. In study of analytic formulation, researches often presumed that lead-time is constant and/or demand quantity and time between two demands are constant or special functions. In study of computer simulations, the retailer's discount to customers isn't considered commonly.To make retailers get largest profit, an order-inventory-sale simulation model is created for single kind of merchandise with shelf-life based on the simulation principle of the discrete events system. The model at one time dissussed order quantity discount given by producer, stochastic lead-time, price discount given by retailer to reduce his inventory quantity, costumers' stochastic demands and stochastic time between demands of customs and the increase in the costumers' demands due to price discount.And the model is extended to solve inventory problem of correlative merchandises. By data experiments, validity of models is analyzed.The simulation models created in this dissertation solve the inventory problem of merchandise with shelf-life successfully and supply retailers with efficient tools to select best order-inventory-sale policy.