Interpretation in-situ test data using artificial neural networks

Establishing a realistic working profile of soil properties has been, and is still, one of the most challenging problems facing geotechnical engineers. In the present study a neural-network approach is used to tackle this problem. Source data of a series of Standard Penetration Tests (SPT) performed at the Texas A&M University's National Geotechnical Experimental Sire are used for training and testing artificial neural networks. The developed neural network is shown able to predict the SPT N-values of the site studied. Data are then generated for constructing the profiles of the N-values using the trained neural network, The study shows that the potential of neural networks in site characterization is significant.