Tunnel Surrounding Rock Parameters Inversion Study Of Support Vector Machine (SVM) Method

The displacement back analysis based on support vector machine (SVM) is a new kind of parameter inversion method in geotechnical engineering, which uses the improved support vector machine (SVM) model to replace the numerical model to establish a nonlinear relationship between mass mechanics parameters and the displacement and greatly improves the efficiency of the inversion calculation. This paper focuses on some defects which currently exist in inverse analysis method, and the rock mass mechanics displacement back analysis method based on support vector machine (SVM) is researched systematically respectively from the advantages and disadvantages of the traditional deterministic displacement back analysis method, the defects and basic principle of support vector machine (SVM), the performance of improved fish algorithm, support vector machine (SVM) optimizated by the improved fish algorithm four aspects, which is applied to the engineering practice. The main research work and conclusions are as follows:(1) The principle and method of displacement back analysis is introduced, and the traditional deterministic method has defects such as large amount of calculation, poor practicability and so on, and some uncertainty methods such as intelligent thought has become the development trend of Parameter inversion, and the possibility of practical application of displacement back analysis is summed up.(2) In view of the shortcomings of neural network such as previous learning, large sample and so on, this paper proposed that support vector machine (SVM) method was used for displacement back analysis. The basic principle of support vector machine (SVM) is introduced systematically, such as the statistical learning theory, VC dimension theory, the definition of the kernel function and the related parameters. The problems existing in the support vector machine (SVM) is pointed out. which is the theory support for the improvement.(3) Because the support vector machine (SVM) uses gaussian kernel function as a general inner product to solve practical problems, but the biggest problem is how to choose the suitable model parameters, such as the width of the gaussian kernel and punishment parameters. In this paper, usethe improved fish algorithm to search all the parameters of support vector machine (SVM), and it improves the fitting efficiency of support vector machine (SVM). The example shows that the improved fish algorithm significantly improves convergence speed and accuracy.(4) Combined with the project example, successfully use FLAC3D finite difference method to calculate and simulate for tunnel surrounding rock parameters inversion, then use the orthogonal experiment design principle to structure the support vector machine learning and prediction data, finally inversion parameters return to the FLAC3D for generation, and the relative error between positive numerical simulation value and actual value is within 3.0%, which verifies the validity and reliability of the approach.(5) Previous work as the basis, combining with the Jiangxiwuji highway tunnel example, engineering application is researched. After the sample data is constructed by FLAC3D and the orthogonal experiment, use support vector machine (SVM) optimized by the improved fish optimization to inverse tunnel surrounding rock mechanics parameters, finally the actual value of the 6 line displacement verify the rationality and effectiveness of this model, which is compared with the forward modeling results. Therefore, the parameter inversion method in geotechnical engineering provides a research idea for the future research.