Application And Study Of Support Vector Machine In The Underground Engineering

Support Vector Machine is a new pattern recognition technology which is based on VC Dimension theory of Statistical Learning Theory and Structural Risk Minimization of it. SVM can obtain the optimum result from the information having been gained which is not only the optimum result when the samples are infinite and has much stronger theory and better generalization than Neural Network which is based on Empirical Risk Minimization; The algorithm is a quadratic programming problem which can ensure the extreme result is the globally optimum result and it solve the inevitable local extremum problem of some Artificial Intelligence Methods. The algorithm solve the practically problem in the high-dimension feature space by non-linear transform and construct linear distinguished function in the high-dimension space to finish the non-linear distinguished function in the former space which can ensure the better promotion capacity of learning machine and stronger non-linear construct model; Also, it solve the Curse -of-Dimensionality problem .This paper applied SVM to the underground engineering , because it has better learning feature and future apply value which is hoped to solve much problem in the underground engineering. The following work was conducted in this paper:The first work simply introduced Statistic Learning Theory of the theoretic base of SVM.The second work detailedly deduced the training and decision-making process of SVM from linear SVM to Non-linear SVM and sum the training algorithm.The third work constructed the regression model on the base of theory of SVM ,then applied this model to the prediction of vault sink displacement detecting data of YK37+215 section of HuaYingshan Tunnel and the regression analysis of the fractured zone of the HuaFeng coal roadway walls by programming in MATLAB language.The fourth work put forward rock mass classification algorithm based on SVM and constructed the rock mass classification model based on SVM, then succeeded applying this model to the second project surrounding rock classification of GuangZhou pump accumulator electricity station and synthesis three-character rock classification.In conclusion, applying SVM to rock mass classification and underground engineering regression-analysis of testing data is feasible , also it has need less learning sample, more prediction-precision , stronger performance of dealing with non-linear dynamic data and better performance of non-linear system modeling than other Artificial Intelligence methods and has a advancing promise.