Estimation of Unconfined Compressive Strength by Spatial Interpolation Using Non-Geostatistical Methods and Artificial Neural Networks

This study applies spatial interpolation to estimate soil engineering properties by using previous information in the neighborhood areas. This study focuses on soft clayey Bangkok soil data in the Bangkok Thailand. The non-geostatistic and artificial neural networks (ANN) methods are compared to estimate unconfined compressive strength of soil. The non-geostatistics are inverse distance weighted, triangulation, natural neighbor, b-spline approximation, cubic spline approximation, global thin plate spline, local thin plate spline and thin plate spline. For this study, ANN is the four layers feed forward neural networks with error back-propagation learning. From the computation with the testing data, the cubic spline approximation gives the lowest RMSE. ANN is also applicable with more input data.