Prediction of TBM boring speed based on IPSO-LSSVM parameter optimization algorithm

Scientific prediction of TBM tunneling efficiency has important guiding value and significance for long-distance tunnel construction risk and project cost control. A TBM tunneling speed prediction model based on the least squares support vector machine algorithm optimized by the improved particle swarm optimization algorithm is proposed. The standard particle swarm optimization algorithm is improved by using adaptive inertia weight, and the regularization parameters and kernel parameters of LSSVM are optimized based on the improved PSO algorithm. The tunneling speed prediction of TBM is trained by using the tunnel data of No.3 tunnel in Queens, New York and Karagyi-Tehran diversion tunnel in Iran. After the training reaches a certain accuracy, the actual engineering data of TBM construction of long tunnel in Xinjiang are predicted. The results show that the proposed IPSO-LSSVM model has a prediction fitting degree of 0.9776 for TBM tunneling speed and an average absolute error of 1.0771%, which is more accurate than the prediction results of SVR, LSSVM and PSO-LSSVM models. The prediction error does not exceed 1.5% in practical engineering.