Study On Location Selection Of Automotive Components Distribution Center Based On Genetic Algorithm

With the development of electronic commerce, as its core industry, logistics distribution has unconsciously played an indispensable role in people’s life, and distribution centers is to maintain an important part of the system can run properly, the essential role. Building logistics distribution centers scientifically can significantly reduce supply chain costs, associated companies and users in order to protect the overall effectiveness of the implementation. Therefore, the logistics distribution center location in raising the overall efficiency and effectiveness of logistics is of great significance.To distribution center location problem, traditional algorithms have many limitations, the genetic algorithm using a computer to calculate the optimal location strategy, however, the algorithm is still in its infancy, and in order to solve practical problems, many parameters need to be optimized. Traditional genetic algorithm suitable for solution space is small, and the limited number of solution and discrete conditions. In the case of the solution space is larger, number of solutions are infinite, or the distribution of solutions are continuous, result in calculation have soared, moreover is not easy to get the optimal solution. In the case of solution space is larger, numerous scholars put forward many improved strategy, based on predecessors’ results, this paper has carried out the following research:Firstly, the analysis and modeling has been carried on distribution center location using Baumol-Wolfe. Case study analyzed the logistics marshalling data of car factories that located in north China for five years, based on the data, two different models are presented. According to models, three kinds of genetic algorithm are proposed, then one or more distribution centers were selected to find the best set of distribution center location selection strategy in order to minimize the total cost of the supply chain by genetic algorithm.Secondly, this paper analyzes the location problem,the goods-can’t-be-split model and the goods-can-be-split model are proposed. Firstly this paper introduces the Baumol-Wolfe model. Because Baumol-Wolfe method is easy to fall into local optimal value, the first genetic algorithm is proposed. It is based on goods-can’t-be-split model, and its solution space is limited. For the goods-can-be-split model, the solution space is infinite, data is continuous and big.so we uses the penalty function in the second genetic algorithm. What’s more, we improve this algorithm, such as improving the initial population, the best individual reservations when select, jump mechanism when falling into local optimal value. We obtain acceptable results. However, the second genetic algorithm is still not good enough. We presents a constraints substituting method to reduce the search space further. This is the third genetic algorithm. Satisfactory results are obtained. At the same time, I discussed the parameters of genetic algorithm such as mutation rate, the role of crossover rate in actual programming and how to select these parameters to get better results and improve the capability of programming and debugging.Finally, the scopes of application, accuracy and precision of three genetic algorithm are compared. And we discuss how to use genetic algorithm in different situations, it provides a detailed theoretical basis for future research.