Comparative Research Of The Ant Colony Optimization’s Parameters Setting For Tree Pipe Network Optimal Layout Based On The Uniform And Orthogonal Experimental Method

In the people’s daily life and social economic stability development,Urban water system,one important municipal infrastructure,plays an irreplaceable role.Water distribution network for the water supply engineering investment usually take over around 60%in the water supply system project construction.Using the optimal theory and computer network technology,optimum arrangement on pipeline network and the optimization design can improve the engineering benefit,reduce cost of the investment,project construction and the follow-up for operation and management of important data reference.On the basis of former,it’s always an mportant problem in the technical personnel field that exploration and expansion and new algorithm are applied to the water supply engineering.Ant Colony algorithm(Ant Colony Optimization,hereinafter referred to as ACO)is proposed by Italian scholars M.D origo in his doctoral dissertation in 1992,which is successfully applied to different fields and gain valuable experience.It’s a combination of solving Optimization problem new heuristic method that has the positive feedback,distributed computing and abound constructive greed heuristic search characteristics.In this article,we research ant colony algorithm using in the tree network.According to difficulting to set up the parameters of the ant colony algorithm,we use uniform experiment method and the orthogonal experiment method to optimize the parameters for exploring an new algorithm to determine the parameters.The paper mainly research content and conclusion includes the following respects:(1)It’s feasible to use ant colony algorithm solving the arrangement about tree pipe network.This article based on the total length of pipe network shortest,with small path choice of water supply network as an example,use the method of ant colony on tree network layout optimization.After testing to verify the application,it is feasible ant colony algorithm using in the tree pipe network optimum arrangement,and this method can achieve a prospective result.Compare with sigle genetic algorithm,the optimization of the parameters which is not optimized of ant colony algorithm can get tree network,but the relatively slow speed,which explain that the choice of parameters of ant colony algorithm to optimize performance impact needing to further the parameters optimization algorithm.(2)The main parameters of ant colony algorithm include the population of ant,Alpha,Beta,Rho and QT.First of all QT equals to 5 when the population of ant,Alpha,Beta,Rho are the same situation,ant colony algorithm is stable and the output of the total length of the minimum spanning tree fluctuate minimum.In QT for 5 cases,in order to improve the running time of the ant colony algorithm for the purpose,we choose single index of uniform design of experiment method to the population,Alpha,Beta and Rho four parameters optimization of 9 level test,and use the intuitive analysis result analysis.Research shows that even simulation test results indicated that the optimum experimental conditions of Alpha = 1.7,Beta=5.0,Rho = 0.56,m = 5,the time is 2.9 s.In the tree pipe network layout,the ant colony algorithm performance got great improvement,shortened running time,improved the operation stability,and could effectively avoid the parting optimal.(3)At the same situation of QT=5,as the index which is in order to improve the running time of the ant colony algorithm,the single index orthogonal design method for the population m,Alpha,Beta and Rho fourparameters for the 3 level of optimization experiment,and by using the intuitive analysis and a poor analysis result analysis,the parameters set combination for Alpha = 1.5,Beta = 4.2,Rho = 0.65 and m= 9,running time=3.0132 s.In the research,when the runing time is same,although even test method and the orthogonal experiment method to determine the parameters of the different combination,the average tree and the shortest length tree are greatly improved based on the uniform test method of the parameters.If using the orthogonal experimental method,although test times has decreased,parameter combination is not very thorough..Therefore,uniform experiment method is used to determine the parameters of the ant colony algorithm optimized combination than used orthogonal experiment to determine the optimal parameters of ant colony algorithm.(4)In the situation of a uniform experimental method to get the best para meters combination,ant=5,Alpha=1.7,Beta=5.0,Rho=0.56,QT=5,ant colony alg orithm,whose the parameters is optimized,is compared with genetic algorithm.Research shows that the ant colony algorithm through the parameters optimizati on,can get minimum tree,and can effectively shorten the running time,get tim e only t=3.0219s for the optimization results.Research shows that it is feasible for the ant colony algorithm to apply tr ee pipe network optimum arrangement.Both the uniform test method and the o rthogonal experiment can undertake ant colony algorithm for parameters optimi zation.Relatively speaking,based on the uniform experimental method is used to determine the parameters of the ant colony algorithm combined more fast.Fo r tree pipe network optimization layout problem,compare with the single geneti c algorithm performance,the running time of ant colony algorithm is more sh orter.But ant colony algorithm which gets the optimal solution of the stability remains to be further improved,and still need to further improve the algorithm,test the mass disposal of pipes in the application effect.