Reliability Analysis Of Large Span Cable-stayed Bridge Based On Support Vector Machine

This paper summarized the existing structure reliability theory and calculation method,summarizes the advantages and shortcomings of each method,using the new support vector machine response surface method to fitting functions of long-span cable-stayed bridge structure,and choose the appropriate intelligent algorithm to solve the reliability.The main research contents are as follows:(1)This paper sums up the shortcomings of the traditional reliability method: the FORM method and SORM method needs to know the expression of function,and the accuracy is low when the function is complex.Monte carlo method is very computationally intensive;Two polynomial response surface method is less accurate when dealing with highly complex functions,and the neural network response surface method is easy to be understudied,overstudy and local optimal.It is found that the support vector machine algorithm has the advantages of low fitting accuracy,strong learning ability of small sample,and the best overall result,and it can get the function expression that conforms to the force performance of the cable-stayed bridge.(2)In standard positive space,reliability is the shortest distance between the limit state surface and the origin of coordinates,which is a constrained optimization problem.The penalty function method can be used to transform it into the minimum value problem of unconstrained function.Intelligent optimization algorithm can solve these problems better.Selecting the artificial fish algorithm,genetic algorithm,particle swarm optimization algorithm which are often used in the function extreme value problem to Comparison and analysis their effectiveness,found that the calculation accuracy and stability of the Particle swarm optimization algorithm is best and it is more suitable for solving the reliability.(3)The accuracy and validity of the algorithm are verified by two examples.This article tells the reliability calculation method by matlab programming platform,which can greatly reduce the amount of calculation and significantly improve the interface environment.the reliability of the structure and other engineering can be solved as well,the program has certain practical value.(4)It is found that the reliability of the bridge structure is high and the safety is very good by calculating the reliability of the span deflection,cable stress and main tower stress of the large span cable-stayed bridge.This paper analysis the parameter sensitivity of randomvariables which affects cable-stayed bridge reliability,studies how the mean and standard deviation of the random variable change can influence reliability,gets the relationship between reliability and the mean and standard deviation of the random,obtains the influence degree of each random variable on reliability.The conclusion has great reference value in the design,construction and management of cable-stayed bridge.At last,the reliability of the cable-stayed bridge system is calculated,and the reliability index is high.