Back Analysis of Tunnel Displacements Based on the IAF-SVM Algorithm
(1 Civil Engineering College of University of Science and Technology Liaoning, Anshan114051; 2 Road Administration of Hebei Province Expressway Management Bureau, Shijiazhuang 067512)
Abstract:
In light of the defects of overtraining a sample of a BP neural network and the low precision of a small sample, a SVM-based back analysis method for tunnel elastic-plastic displacement is proposed using the generalization ability of a support vector machine(SVM). Considering that the performance of the support vector machine (SVM) largely depends on the selection of parameters, the efficient global search ability of the improved artificial fish (IAF) is adopted to get the optimal parameters of the SVM to avoid randomness in parameter selection. A tunnel is analyzed using FLAC3D and the inversion of the elastic-plastic displacements is conducted based on the calculated displacements of measured points. The results show that the convergence rate and inversion precision of this method are better than that of the BP neural network regarding small samples.