The Application Study Of Nonlinear Forecast On Surrounding Rock Displacements Based On LS-SVM

In construction of tunnel or underground excavation, forecast of the displacements of surrounding rock according to the actually measured data is significant practical meaningful to support design, construction and dangerous condition predicting.Because surrounding rock system is nonlinear, fuzzy and uncertainty, conventional precise mathematic model has relatively large difference to the practical situation since it is under so many assumptions,and its results of forecast are not as good as request.Support Vector Machine(SVM) is a new kind of machine learning algorithm proposed recently which is based on VC Dimension Theory and Structural Risk Minimization of Statistical Learning Theory. SVM can obtain the optimum result from the gained information which is not the optimum result only when the samples are infinite. SVM has much stronger theory foundation and better generalization than Neural Network which is based on Empirical Risk Minimization. Least Squares Support Vector Machine(LS-SVM) is an improved algorithm to conventional support vector machines.It simplifies the model parameters and speeds the computations by solving a set of linear equations instead of quadratic programming for classical SVM. In this paper, LS-SVM is introduced to the forecast analysis of displacements of the surrounding rock. And the forecast model is set up with LS-SVM to forecast displacements of surrounding rock.Firstly, numerical simulation is carried out by using two different fashions with data generated by one choosed function. One is by regression analysis methods,the other is by time series forecast model with equal step length. And noise is added to generated data to simulate actual environment.Model parameters are obtained through cross validation method that can effectively solve the parameter estimation problem. RBF neural network is applied to time series forecast with the same data in order to compare the forecast effect with LS-SVM model.The simulating experiment demonstrates that the forecasting effect of using LS-SVM model was better than that of using RBF neural network model when time series forecasting with equal step length is performed. Experiment result also indicated that a higher precision was obtained with regression analysis model in forecasting missing data of intervals of training data than in forecasting ahead.And time series forecasting with equal step length has a better effect in forecasting ahead.Combining the advantages of regression analysis methods and time series forecast model with equal step length,a compound forecasting model was set up ,and was tested with engineering data. The forecasted results showed that this model has perfect estimate capabilities.