Prediction of limit pressure and pressuremeter modulus using artificial neural network analysis based on CPTU data

Abstract

Pressuremeter test (PMT) is conducted to obtain effective soil parameters such as limit pressure (P_L) and pressuremeter modulus (E_p) that are frequently used in calculating foundation bearing capacity, settlement, and foundation behavior. However, the application of PMT in China was limited due to higher cost and time. There is a need for identifying a suitable method and establish models to predict reliable P_L and E_p for interpreting or cross-checking soil parameters. Piezocone test (CPTU) offers an ideal test method to develop correlation models since it is widely adopted for geotechnical investigations in China. In this study, artificial neural networks (ANN) have been used to develop CPTU-PMT correlations. A total of 92 sets of sandy soil and 65 sets of clayey soil data from four testing sites were collected using CPTU and PMT. ANN was employed to develop 4 models, half of them considering effective overburden stress (${\sigma}_{v0}^{\hbox{'}}$), for predicting P_L and E_p from CPTU data. The obtained ANN models were validated using the measured values of P_L and E_p from pressuremeter tests and also the predicted values based on previous correlations. The comparison results show that P_L and E_p values predicted by ANN models proposed in this study are more consistent with the measured values at testing sites. Additionally, foundation settlements were measured from a load test and compared with predictive settlements obtained using P_L and E_p estimated by the proposed ANN correlation models. The results have shown that the CPTU results can be used to accurately predict PMT parameters and derive settlements.