基于机器学习的土壤不排水抗剪强度模型研究

土壤抗剪强度在岩土工程中扮演着重要的角色,通常需要结合原位试验和实验室试验来进行测定。为了减少大型工程项目对繁琐且昂贵的实验室试验的依赖,探究了5种机器学习模型(人工神经网络、支持向量机、高斯过程回归、随机森林、XGBoost),利用静力触探试验数据来预测土壤不排水抗剪强度,以评估机器学习技术对减少实验室试验需求的潜力。结果表明,在5种模型中,XGBoost(R^(2)=0.76,MAE=10.79,RMSE=15.45)与ANN(R^(2)=0.76,MAE=10.87,RMSE=15.49)表现最出色,具有预测土壤不排水抗剪强度的潜力。然而,模型在组10折交叉验证中表现不佳,这表明尽管模型在训练时能够很好地处理特定场地数据,但在没有特定场地数据的情况下,泛化能力受到限制,凸显了准确描述土壤异质性的挑战。

Low secondary compressibility and shear strength of Shanghai Clay

In order to investigate the compressibility, particularly the secondary compression behaviour, soil structure and undrained shear strength of Shanghai Clay, a series of one-dimensional consolidation tests （some up to 70 d） and undrained triaxial tests on high-quality intact and reconstituted soil specimens were carried out. Shanghai Clay is a lightly overconsolidated soil （OCR=1.2-1.3） with true cohesion or bonding. Due to the influence of soil structures, the secondary compression index Ca varies significantly with consolidation stress and the maximum value of C~ occurs in the vicinity of preconsolidation stress. Measured coefficients of secondary compression generally fall in the range of 0.2%-0.8% based on which Shanghai Clay can be classified as a soil with low to medium secondary compressibility. The effect of soil structures on the compressibility of Shanghai Clay is found to reduce with an increase in depth. Soil structure has an important influence on initial soil stiffness, but does not appear to affect undrained shear strength significantly. Undrained shear strengths of intact Shanghai Clay from compression tests are approximately 20% higher than those from extension tests.