Intelligent Algorithm In The Application Of Bridge Reliability Assessment

Assessment for Bridge reliability is an important mean for bridge custody. Currently; reliability theory is widely used for its precision and visual expression advantages. Monte Carlo method based on the experts experience is most commonly used among many reliability assessment methods, but it is inefficient, and also has a weak adaptability and excessive depend on experts experience. The project "bridge remote monitoring and reliability assessment", which use the Monte Carlo method, gets accurate results, but its disadvantages are also obvious. In order to apply this method more widely, it is essential to enhance the efficiency of reliability assessment. With the development of intelligent algorithms, such as neural networks and genetic algorithms, it is possible to use these intelligent algorithms to build bridges'reliability assessment model.This paper achieves the following researches which use intelligent algorithms in the bridge reliability assessment:First, a reliability assessment the bridge model based on BP neural network is proposed. This model reflects reliability assessment of the bridge just analyzed data getting from sensor monitors which are arranged in key position of the bridge rather than cumbersome limit state equations for complex structure of large span bridge. Moreover, during the bridge lifetime, the model can be reconstructed to get the dynamic assessment results. According to different force structures, the model can be used in different bridge reliability assessments by adjusting the input layer and the hidden layers of the BP neural network.Second, a Bridge reliability evaluation model based on the real-coded genetic algorithm optimization BP neural network (GA-BP) is addressed. Because of fault-tolerant and global optimal characteristics of genetic algorithms, GA-BP neural network overcomes the vulnerability of local optima and the defect that convergence is much relied on initial weights. Moreover, the traditional binary coding based on genetic algorithm would result in the jumping of sample information, and then may cause errors. To solve this problem, the model uses the real coding which make it more accurate for the reliability of the bridge.Third, using PowerBuilder and SQL Server, A system for bridge reliability evaluation, based on Monte Carlo Law and experts'experience, is present. The system's basis is the bridge reliability assessment application model of real-coded genetic algorithm optimization BP neural network (GA-BP).Finally, the data of MaSangXi Bridge is used to analyze previous models. Experiments show that two new methods are more effective than the Monte Carlo method. They can satisfy the precise requirement of reliability assessment of the bridge project. The bridge reliability assessment model based on the real-coded genetic algorithm optimization BP neural network (GA-BP) is more precise than model just based on BP neural network. Moreover, these models solve the problems of traditional assessment methods, and they can be used more widely in engineering.