The Application Of Fuzzy Clustering And Partial Least Squares For Dam Monitoring Data Analysis

China has a large number of dams and reservoirs. The analysis of dam observation data is an important way to master the real working state of dam, provides a scientific basis for the normal operation of the hydropower station. Therefore, it is necessary to analyze the observation data of dam in time, effectively.Generally there is a large number of security monitoring facilities laid in the dam.Due to the high frequency of monitoring, it often accumulates large amounts of monitoring data. If build a model for analysis by-point, it will be time consuming, and it is not conducive to grasp of the dam running state timely. Therefore, this article by based on the fuzzy relation of clustering analysis quickly gained category deformation observation point, and established a partial least squares regression model for the typical observation points. The main contents and results are as follows:(1) With the one hydropower station as an example, using the clustering method based on fuzzy equivalence relation to make a clustering analysis of deformation monitoring of dam and lock. By use of fuzzy transitive closure method to transformed fuzzy similarity relation into fuzzy equivalence relation. And select threshold alpha for clustering. By use of the method average installments deformation allowable difference between the points to evaluate the result of the classification, select the best classification.(2) To established a partial least squares regression model for the typical observation points. It choose water level, temperature, time as the factors of the regression model. And it effectively solves the problems which are multiple correlations between factors and modeling the less sample size.(3) Using partial least squares regression model to predict deformation data of the remaining five typical measuring point, and comparing with the measured values, the results show that the partial least-squares regression equation established forecast better.(4) By fuzzy clustering analysis of deformation monitoring data processing, obtained the best classification of deformation monitoring. Select a certain number of measuring points in each category to build a mathematical model, and then use this model to analyze this kind of deformation monitoring. The results show that the deformation monitoring points classification reasonable, and elected a typical observation points, reducing the complicated calculations, improve work efficiency of monitoring data analysts.