A support-vector-machine-based method for predicting large-deformation in rock mass

High in situ stress and some geological disasters induced, such as large deformation, are well known as the complex engineering problem which often happen in tunnel construction. Large deformation can bring great threat to tunnel safety during construction and operation, but so far there is no concrete and applicable methods for the prediction of large deformation, it needs to to further developed in practice. In the paper, the forecast methods on the basis of the concept and mechanic of large deformation in surrounding rock were systematically summarized and a new comprehensive method to judge and forecast large deformation was brought up accordingly. Besides the popular criterions, this method adopts the support vector machine (SVM), a new method of machine learning which is applicable to solve the nonlinear problem with smaller samples and better precision than traditional ways, and predicts the deformations of surrounding rock by establishing the nonlinear relationship among deformation time series. Meanwhile, it conducts the emulated mock experiment on the deformation in tunnel construction by using Finite Element Method (FEM) software called ANSYS. The result showed that this new method was feasible and convenient. What´s more, it is superior to traditional methods in efficiency and feasibility, and valuable for the engineering practice.