Research On Identification Of The Damages In Bridge Structure Based On Combining Neural Network And Genetic Algorithm

With the fast development of communication in our country, density of highways network improves constantly and the large span bridge is emerging constantly. But structural damages such as cracking and aging etc. to some degree during the bridge structures' service life occur from time to time. Some invisible damages do not alert the engineers in time and may result in disastrous consequences and ultimately cause substantial loss of life and property. Therefore, it is important to command the health status of bridge in commonly using state in time and to detect damages at the earliest possible stage. The main researching fields are as following:On the basis of analysis of the data about structural damage detection, Artificial Neural Networks(ANN) and Genetic Algorithm (GA), three damage identification methods are summarized, including: static state identification approaches, dynamic identification approaches, and ANN combined with GA aptitudinal identification approaches. The theories, formulations and usages of every approach are discussed systematically.The aptitudinal and predictable method based on combining Artificial Neural Networks (ANN) and Genetic Algorithm (GA) in structural damage detection are proposed. The procedure of identifying damage can be defined as a minimization problem. The optimum solution can be obtained effectively by using GA. Since GA usually needs a long analyzing and calculating process in use with the Finite Element Method (FEM). But a non-linear mapping function from multiple input data (structural damage parameters) to multiple output data (differences of response between damaged structure and intact structure calculated by FEM) is constructed within BP neural networks. The ability of constructing a non-linear mapping function within BP neural networks offer the strong calculation means which solve the problem of identification of the damages within Genetic Algorithm. The method and the approach about identification of the damages combined BP Neural Networks with Genetic Algorithm are given in this paper.The study is based on Hanjing River bridge in Zhongxiang. According to thedesign, construction data and experimental record of the bridge, damage-detection-oriented finite element model of the bridge is established and the analysis is then carried out. To the actual disease of this bridge, the study of damages identification from two aspects such as elastic modulus and pre-stress was investigated. First, the training samples of BP neural network are designed using orthogonal test and the input data of BP neural network consists of damage locations and damage extents. Second, the differences of displacements between damaged structure and intact structure calculated by FEM (GQJS) are made up the output data. Finally, using network trained, the damage parameters of the structure are then detected by the Genetic Algorithm, and the real bearing capacity after damaging in this bridge is estimated with the surveying data. The result indicates that the damage detection method studied in this paper is rational and effectual. The method not only offers reliable scientific basis for making the strengthening scheme of the bridge, but also can be popularized and applied to strengthen and optimize analyzing for other disease bridge similar to the bridge.In this paper, the damages in bridge structures are simulated and analyzed. The method of the damages identification in pre-stressed continuous frame bridge structures based on combining Artificial Neural Networks(ANN) and Genetic Algorithm(GA) is applied successfully. The result shows the effectiveness of this method used in long-span bridge structures.