Prediction of the Consolidation Coefficient of Soft Soil Based on Machine Learning Models

Abstract

The coefficient of consolidation (Cv) of soft soil is a parameter that reflects the consolidation characteristics of soil under load, but it usually costs a lot in time and money to test. In this paper, an artificial neural network (ANN) and a support vector machine (SVM) are used to establish the Cv prediction model, and the soft soil data of Guigang Beihai Expressway in Guangxi are used to train and test the model. Eleven physical and mechanical parameters of soft soil are statistically analyzed by correlation matrix. Four parameters are determined as input parameters of the calculation model to train and test the calculation model, and the performance and robustness of the prediction model are checked. The results show that the ANN model and SVM model both accurately calculate the Cv, with coefficient of correlation R² > 0.91, root mean square error RMSE < 0.2079 cm²/1000s, and variance ratio VAF > 90%. The prediction accuracy of the ANN model is better than that of the SVM model, and the Monte Carlo simulation results show that the SVM model is most robust. Therefore, the consolidation coefficient is connected with other physical and mechanical parameters, and the ANN model and SVM model are used to predict the Cv, which provides a new idea for fast calculation of the Cv.