Research On Displacement Deformation Prediction Of Tunnel Surrounding Rock Based On Support Vcetor Machine

NATM tunnel construction method is widely used. Monitoring and Measurement NATM is an important component part of the measured data can reflect the state of tunnel rock deformation, relying on monitoring data to predict the deformation of tunnel surrounding rock can be a good reference value for the design and construction of the tunnel nowadays displacement back analysis of the tunnel can be predicted within a certain range, but the inversion analysis mainly rely on the material properties of the stable value of the settlement with the neighboring interface inverse similarities, there are some lag can not meet construction projects time invested on request. Self regular use of measured data to predict the deformation of the displacement is also less. For time series prediction displacement, the paper relies on statistical learning theory of support vector machine to carry out research, the main work are:① Introduces the theoretical basis of support vector machine(SVM) and the derivation process of the concrete, using Yang Zong example analysis of the displacement monitoring data of the tunnel under the different super parameter combination performance of support vector machine(SVM) and compared the polynomial regression prediction and the difference of the recursive least squares method to predict the results and relying on the principle of using MATLAB programming calculations derived analysis showed that the support vector machine has good stability. show that support vector machine has good stability.② Hot springs in beijing-zhuhai expressway tunnel displacement of the nonlinear time series prediction as an example, compares the different characteristics under the form of RBF kernel support vector machine(SVM) to predict performance. The analysis results show that the data pretreatment can effectively improve RBF kernel function of support vector machine(SVM) to predict performance, for the same subset selection mapping space by RBF kernel function is not sensitive.③ Aim at the characteristic of RBF kernel function is good, through the wavelet theory and the characteristics of the kernel function is constructed Morlet wavelet kernel function, and through the xiluodu power station tunnel monitoring data analysis of the test, Morlet wavelet kernel function in different displacement deformation forecast analysis, monitoring frequency has better prediction performance.