(1. Key Laboratory of Ministry of Education for Geomechanics and Embankment Engineering，Hohai University，Nanjing，Jiangsu 210098，China；2. Research Institute of Geotechnical Engineering，Hohai University，Nanjing，Jiangsu 210098，China； 3. Hydrochina Chengdu Engineering Corporation，Chengdu，Sichuan 610072，China)
Abstract：It is difficult to describe the complex nonlinear relationship between all kinds of geological factors of rock and their mechanical behaviors. A new model for forecasting the mechanical behaviors of rock is proposed by combining the particle swarm optimization(PSO) and the support vector machines(SVM)，which is support vector machine based on particle swarm optimization(PSO-SVM). The model，on one hand，uses the nonlinear characteristics of SVM to establish the nonlinear relationship between geological factors of rock and their mechanical behaviors. On the other hand，the penalty factor and kernel function parameter of SVM are optimized by PSO，by which the accuracy of the parameters used in the model is ensured as well as the precision of forecasting result. The model is applied to forecast the coefficient of compressibility of rock and the result is compared with that of back propagation neural network(BP-NN). It is shown that the forecasting precision of PSO-SVM is higher than that of BP-NN，which indicates that the model here is feasible and effective.
