Structural Health Monitoring Of Grand Bridge Based On Neural Network

Bridges are key elements in the transportation system. With the vigorous development of the transportation industry in our country, more and more bridges appear. However, influenced by factors such as environmental erosion, material aging, and over use, the damages of the bridge become more and more serious, and even cause some major accidents such as bridge collapse, largely threatening people's life and property security. To prevent these accidents and diagnose the status of the bridges effectively, it is necessary to grasp the health information of the bridge while it is operating in time and provide the maintenance and evaluation with scientific and reliable evidence, which is an important project for bridge engineering.In order to identify and position the potential damages of the bridge early, it is necessary to take long-term health monitoring and regular evaluations to some major bridges. During the service period, the detection systems of many large bridges collected abundant deflection data, which can reflect the degree that the observation points deviate from the equilibrium position. For bridges like rigid frame bridge, the deflection data can reflect bridge damage effectively. The object of this thesis is the deflection of Ma Sangxi Bridge in Chongqing, which is obtained from the sensor set up on the bridge.There is no complete set of theoretical guidance for the health detection of the bridge, so no method can provide the absolutely accurate identification results individually. Therefore, after comparing with a variety of computer approaches, this thesis integrated artificial neural network and system-level fault diagnosis and tried to detect the possible damage and hidden danger of the bridge from deflection data. The main contributions of the thesis are organized as follows:Firstly, the thesis summarizes the main approaches and current situation of the health detection of the bridge, and introduces relevant knowledge of the artificial neural network.Secondly, the thesis introduces a new bridge health detection method, which is inspired by the concept of"deflection resonance"and adopts relevant knowledge of BP neural network and system-level fault diagnosis. It also introduces the concept of point, which is composed of a deflection checkpoint and the dedicated sensor. This method consists of three stages. First, establish test pattern between among checkpoints with BP neural network technique. At the second stage, a set of tests on points are conducted based on the established test pattern, and the test outcomes are collected to form a syndrome. Finally, a test result is obtained based on the syndrome. This is the first time to integrate the"deflection resonance"phenomenon with the neural network in the bridge detection system.Thirdly, adopt above method to the simulation detection of Ma Sangxi Bridge in Chongqing and obtain relatively good results, which demonstrate that this method has certain theoretical and potential application values.Generally, this thesis introduces a bridge health detect method based on neural network and applies it to the simulation detection of Ma Sangxi Bridge successfully.