Research On Genetic Algorithms Used In Pavement Management System

Since 1980s,the construction of high-grade highway has developed rapidly in our country. With the Initially completed of highway net, maintenance management work is heavy more and more. In foreign country, PMS have been approved to be a effective tool used in maintenance highway net and give overall planning in fund allocation. Pavement management system is introduced into our country in 7^th Five Years Plan, but research and application on high PMS star soon. This paper research on algorithm used in highway PMS .This paper involves 2 aspects :1,Used Genetic Algorithm in highway roughness combination forecast model.2,The research of pavement maintenance decision optimization basing on genetic algorithms.Since the first time it was put forward in 1969, combination forecast has made great development. Theory research and practical appliance have indicated that combination forecast has better precision than individual ones. This paper presents 2 methods for pavement roughness combination forecast, one is linear pavement roughness combination forecast, which used genetic algorithm to compute the weight for each single pavement roughness forecast model. And another is non-linear roughness combination forecast, in this method , genetic algorithm is used to design the framework ,threshold and weight of network that makes the training process tending to global optima. Satisfied chromosomes can be found after choosing, acrossing and aberrancing.There are two optimization solutions to maintenance decision, one is the math programming, and the other is the artificial intelligence which mostly refers to genetic algorithms. Though the math programming is used wildly in PMS domestically, the solutions are low precise and used a lot of time when the problem is complex at the network level. It is found that the robust search characteristics and parallel handling capability of genetic algorithms are well suited for pavement maintenance decision optimization at the network level.Firstly, this paper introduced the single goal optimal model of pavement maintenance decision and the multiobjective optimal model of pavement maintenance decision are established after analyzing all kinds of pavement maintenance decision models and its optimization approaches at present. Then a hybrid genetic algorithm is proposed to solve the single goal optimal model of maintenance decision basing on analyzing the advantage and disadvantage of simple genetic algorithms. The hybrid genetic algorithm with characteristics of adaptive pseudo-parallel and optimum maintain is capable of overcoming premature convergence and find global optima efficiently. Moreover, It is proves to be effective through solving the single goal optimal problem of maintenance decision. At last, the current state of the research. on the basic theory of genetic algorithms about multiobjective optimization is systematically presented. In addition, great importance is attached to the introduction of the elitist nondominated sorting genetic algorithm (NSGA-II), which is widely used in the problems of multiobjective optimization. At the same time, this method is provedto be more effective than math programming after the multiobjective optimal model of maintenance decision optimized by NSGA-II.