Abstract:
This study presents the application of artificial neural network (ANN) system for predicting the
undrained shear strength of Bangkok clay. In this study, the program which developed from ANN is
trained to understand the relationship of soil's basic properties using 3 sets of information: Set 1,
contains 3 parameters which are depth, moisture content and unit weight. For set 2, it composes of 4
parameters which are depth, moisture content, unit weight and SPT. The final set constitutes 6
factors which are depth, moisture content, unit weight, Liquid Limit, Plastic Limit and Plasticity
Index. These parameters are studied in order to find the relationship between them and to find
functions that can appropriately approximate the undrained shear strength. Then, the undrained
shear strengths are predicted by the program and compared with the undrained shear strengths from
the actual testing for evaluating its accuracy with different types of ANN. The results illustrate that
the relationship which contains 6 parameter (hidden layer 1 layer 25 units) is suitable and the
appropriate function is tansigmoid with 1000 iterations having R- of training of 0.990. On the other
hand, R2 of testing is 0.8290. Moreover, in comparing between the neural network and common
programs, the neural network is more precise than the others. Hence, the neutral network system is
another option so as to help civil engineers predict and monitor the undrained shear strength by
using only physical property tests.