基于CPTu数据的黏性土不排水抗剪强度#br# 神经网络预测

利用人工神经网络模型，建立基于孔压静力触探（CPTu）现场测试数据的黏性土不排水抗剪强度的预测方法。为建立和验证人工神经网络模型，在3个场地开展CPTu和十字板剪切现场测试，共取得33个测孔的CPTu试验数据和相对应的不排水抗剪强度实测值。通过对比分析不同输入向量、不同网络隐层数、不同神经元数及不同改进算法对人工神经网络模型性能的影响，确定人工神经网络模型的具体形式。通过对训练组数据开展机器学习，所建立的人工神经网络模型能够有效地基于CPTu获得的端阻力和孔隙水压力现场测试数据对黏土不排水抗剪强度进行预测，预测结果与十字板剪切试验实测结果非常接近。与传统用于估算不排水强度的经验关系相比，采用人工神经网络模型预测结果与实测结果相关性显著提高、误差明显降低。

Prediction of the undrained shear strength of clay from CPTu data using artificial neural network

A method for the prediction of the undrained shear strength of clay based on CPTu data is developed using an artificial neural network model. In-situ CPTu and vane shear tests are conducted on three sites to acquire CPTu and the corresponding undrained shear strength data from 33 locations, which are used to establish and validate the artificial neural network model. The influence of input data dimension, hidden layer number, neurons in hidden layers, and improved training algorithm on the prediction error and stability of the model is analyzed. Based on the analysis results, a model structure is chosen for the prediction of undrained shear strength. After machine learning through the training set, the proposed model is able to effectively predict the undrained shear strength of clay based on the tip resistance and pore water pressure obtained from CPTu data. The predictions are in good agreement with the vane shear test results. Compared with traditional empirical undrained shear strength estimation equations, the proposed artificial neural network model can significantly improve the correlation and reduce the error between predicted and measured values.