Intelligent Prediction of the Horizontal Deformation During the Excavation Process Based on Particle Swarm Optimisation and Support Vector Machine

The reasonable selection of soil layer parameters relates to the accurate prediction of the horizontal deformation of the foundation pit, which is the main problem of highway tunnel pit design. The aim of this paper is to obtain suitable soil layer parameters for finite element simulation of highway tunnel based on the particle swarm optimisation (PSO) and support vector machine (SVM). First, considering the overfitting problem of SVM in the inversion of soil parameters, the PSO was used to improve the SVM model. Second, the PSO- SVM model was trained with 25 groups of elastic modulus as input values and deformation as output values. Then, according to the monitored deformation data, the soil parameters were inverted by PSO-SVM model. Finally, the inversion parameters were substituted into the finite element model to predict the horizontal deformation of the foundation pit. The results showed that based on the inversion parameters of PSO-SVM model, the finite element method had a good accuracy in predicting the horizontal deformation of the foundation pit. The average relative error between the predicted value and monitored value was 2.95%. Therefore, the application of the parameter inversion method based on PSO-SVM had a reference value for tunnel pit design.