Application of artificial neural network to soft ground improvement: a case study
RIS ID
73702
Abstract
The potential of using Artificial Neural Network (ANN) for the development of a predictive model in soft ground improved by dry deep soil mixing (DSM) is demonstrated in this case study. Previous studies on the application of ANN techniques to DSM data have been limited to the prediction of the unconfined compressive strength (UeS) of soil-cement mix in both laboratory and field conditions. This case study attempts to explore the capability of ANN models to predict field settlements on soft soils improved by DSM using a successive ANN modelling technique. In successive ANN modelling the output of the first ANN model, part of which is the ues, is fed into the second ANN model to generate the settlement predictions. Prior to developing the ANN models, the large amount of relevant information from the Ballina Bypass Alliance project is first organized by categorising into different variables used in soil mix technologies. The data are then integrated into a framework that is acceptable to ANN modelling. A portion of the data is used to train and validate the ANN models; the rest of the data, which have not been introduced into the ANN models, is then used to test the predicition capability of the proposed models. Results of this case study show that the first ANN model is capable of providing reasonable unconfined compressive strength predictions of DSM columns installed in the field. The second ANN model indicates an even better predictive capability when compared to the available field settlement data.
Publication Details
Trani, L. D. (2013). Application of artificial neural network to soft ground improvement: a case study. In B. Indraratna, C. Rujikiatkamjorn & J. S. Vinod (Eds.), Proceedings of the international conference on ground improvement and ground control (pp. 465-470). Singapore: Research Publishing.