关键词: 软土, 固结系数, 人工神经网络, 主成分分析, 鲁棒性

Abstract:

The consolidation coefficient C_v of soft soil is an important parameter in geotechnical engineering. The artificial neural network (ANN) model is improved with the biogeography-based optimization algorithm (BBO). And the artificial neural network with biogeography-based optimization (ANN-BBO) model, trained and tested by using the subgrade soft soil data of the reconstruction and expansion project of Lianyungang–Huai’an expressway, has been adopted to calculate the soft soil consolidation coefficient. Using the correlation coefficient matrix and principal component analysis, eleven physical and mechanical parameters were statistically analyzed, seven of which were identified as input parameters of calculation model which was then trained and tested. The model was tested by correlation coefficient, root mean square error, and variance ratio, whose robustness was analyzed by using Monte Carlo simulation. The results show that the ANN-BBO model can be used to calculate the consolidation coefficient of soft soil, the correlation coefficient R ² = 0.947 1, the root mean square error RMSE = 0.165 7×10⁻³ cm²/s, and the variance ratio VAF = 94.54%. The ANN–BBO model has significantly higher prediction accuracy and better robustness, compared with the ANN model.

Key words: soft soil, consolidation coefficient, artificial neural network, principal component analysis, robustness