Prediction Of Surrounding Rock Deformation For Underground Excavations By Support Vector Machine

ABSTRACT:In underground engineering construction, Deformation of the Surrounding Rock is an important factor in the safety of the underground engineering. surrounding rock instability and collapsing is the most typical engineering accident in underground engineering construction. At present, surrounding rock deformation monitoring occupy an important position in underground engineering. Therefore, the Prediction of surrounding rock deformation becomes a major research direction in underground engineering. Surrounding rock deformation monitoring methods include Empirical Coefficient Method, Safety Factor Method and Numerical Analysis Method. In underground engineering construction, Numerical Analysis Method like finite element method and finite difference method is fairly common. But for underground engineering, Due to the complexity and unpredictability of Surrounding Rock, the selection of constitutive model and its parameters as well as calculation method is subjectivity, makes the Numerical Analysis Method only can be used as the theoretical reference for practical underground engineering. For the reasons above, some scholars have introduced artificial intelligence methods to underground engineering. Predict surrounding rock deformation through establishing mathematical model.Base on previous study, combining Immune Algorithm and Support Vector Machine, also with cross-validation and combined parameter prediction, Predict surrounding rock deformation in underground engineering. The primary studies are as following:(1) Discusses theory and applications of SVM, according to RBV-SVM model, Research on the influence law of SVM regressive properties affected by penalty factor, width of radial basis vector and insensitive factor. SVM play best performance When the parameters achieve the most optimal combination.(2) Establish support vector machine parameter optimization objective function using cross-validation method. Then solve the function and get SVM optimization parameters by immune algorithm which is both search quickly and get the global optimal solution. A predict software Based on support vector machine is developed with MATLAB, and test the effect of the program in Back Analysis and displacement prediction in geotechnical engineering. From the actual application result, SVM with those optimization methods can get a better regression result.(3) Based on the relationship between displacement, axis Force and Time, achieve combined parameter prediction by two-times prediction. Predict displacement using single time series SVM model and combined parameter SVM model and compare the prediction effect of the two. By comparing predictions of the two, the accuracy of combined parameter model is much better.