Chapter Eighteen - Implementing an ANN model and relative importance for predicting the under drained shear strength of fine-grained soil

Abstract

To predict the undrained shear strength (Su) for the fine-grained soil has been following the cone penetration test (CPT) and pore pressure (u) measurements, even though the assessment of strength from CPT results in fine-grained soil is mostly empirical. In fact, for a specific soil, the measured shear strength from the in situ test and laboratory tests may give different results for various reasons. Hence, in this research, the database of in situ and laboratory test data was collected from the literature to predict the undrained shear strength of the fine-grained soils using artificial neural networks. To predict the undrained shear strength (Su) of the fine-grained soils: cone point resistance (Cp), cone side resistance (Cs), pore water pressure (u), and plasticity index (PI) were used as input parameters. To see the accuracy of the predicted undrained shear strength (Su), performance measures were used. The predicted undrained shear strength (Su) from the ANN model was compared with the actual undrained shear strength data having a R2 as 0.98 in both training and testing. Finally, the relative importance is performed and reveals that the cone point resistance (Cp) is the most influencing input parameter having an influence percentage as 45.46.