Computation Of Structural System Reliability Of Long-span Rigid Form Bridge Based On The Support Vector Machine

Continuous rigid frame bridges are used widely in China’s central and western mountains areas, because of its superior structure features and economic advantages. Using the deterministic finite element method to cope with the main security analysis of bridge structures is not enough. In reality, because of lots of random factors in practical projects reliability and probability theories should be used in the main security analysis of the rigid frame bridges. There are many problems which are needed to solve, such as reducing the number of input samples, improving the computation efficiency of calculating the reliability index, the analysis of the system reliability, and the research on sensitivity of parameter distribution in reliability analysis of the bridge structure.Support vector machines is a superior machine learning method based on statistical learning theory, At present, classic reliability calculation method based on SVM has been proposed. By changing the sample data is to achieve the goal of reducing the number of samples. Namely the samples would close to surface by changing the data of load. And the computation efficiency improved, the number of input samples reduced. The reliability calculation method is viable and efficient by comparing other method in examples.Combining with the failure criterion of continuous rigid frame bridge system, this article is trying to determine the bridge control section as well as the main failure modes, then use the proposed method to analysis the system reliability and got the approximate calculation of the scope of the bridge system reliability index by using the method of bound algorithm. At last, sensitivity analysis have been made on parameter distribution on this bridge, including net only the failure mode under the ultimate limit state with normal calculating method, but also the failure mode of excessive deflection in mid-span with a special method when the correlation coefficient of parameters is unknown. On this basis a few construction parameters should be controlled to improve the construction quality and the reliability index during the period of service.