Forecasting peak acceleration of blasting vibration of rock mass based on PSO-SVM

Along with rock mass blasting engineer increasing in our country, how to analyze and predict the peak acceleration of blasting vibration of rock mass from monitor data becomes a focus problem. The monitor data of peak acceleration has complex nonlinear characteristic, which makes it difficult to be analyzed and predicted. This paper uses support vector machine based on statistical learning theory to fit the monitoring data, using the particle swarm optimization to optimize the parameters of support vector machine model, then the nonlinear prediction model of peak acceleration of blasting vibration of rock mass based on pso-svm is constructed. Because support vector machine follows structure risk minimization principle, it overcomes the extra-learning problem of ANN. Because of the rapidly searching optimal parameters by PSO and effectively fetching up the insufficiency of SVM theory, it avoids the human blindness of model parameters selection and improves the accuracy of predictive model. The principle and steps of the method are discussed in the paper. Comparing the monitoring and predicted data of the Tanglang Mountain engineer of Qinshan nuclear power, the prediction model shows good fitting capability. The prediction model offers new analytical method for the peak acceleration of blasting vibration of rock mass.