Building Bridge Health Monitoring Data Mining Model Based On Fuzzy Kohonen Clustering Algorithm

In recently years, our country’s infrastructure was developing fast, bridge structures are important transportation infrastructures of our country, so their security issues are attention of people. In order to ensure the safe operation of the bridge, bridge health monitoring systems are applied widely. The bridge health monitoring system includes a variety of sensors, which collect different data long, real time and ongoing. With the growth of the monitoring time, that would generate huge amounts of data. This dissertation studies data mining base on kohonen algorithm, analysis and calculates the huge amounts of data, then provides the basic data for bridges’ Post-assessment and warning. The main study content includes:Firstly, the algorithm is tests through three random arrays. We analysis the defect of the algorithm based on the clustering results by three arrays. For lack of both convergence speed and clustering accuracy, improvements are proposed. Comparing with original algorithm by three random arrays, we prove the improved algorithm has a good effect.Secondly, using Midas Civil soft to establish the initial finite element model of the bridge according to design drawings of cable-stayed bridge, and a preliminary analysis of the bridge is finished. Through analysis, we know three errors may exist in the bridge finite element model and analysis the cause of each error. In order to obtain bridge baseline finite element model, the initial finite element model can be modified. We modify the model based on dynamic characteristics of the structure and achieve good result. Using the modified model tests the static errors of bridge structure by simulating the load of bridge tests. Through the results of comparing, we prove the error of modified model is less than the original model.Thirdly, for the original data of health monitoring, we conduct the data preprocessing of filtering. Combining kohonen algorithm and baseline finite element model, we build a data mining clustering model of bridge health monitoring and get the threshold of abnormal data mining. Finally, the effectiveness of the model was proved.