Future challenges for artifical neural network modelling in geotechnical engineering

Abstract

Artificial neural networks (ANNs) are a form of artificial intelligence and, since the mid-1990s, ANNbased models have been successfully applied to virtually every problem in geotechnical engineering. This paper briefly examines the areas of geotechnical engineering to which ANNs have been applied, provides a brief overview of the operation of ANN models, and highlights and discusses four important issues which require further attention in the future. These are model robustness, transparency and knowledge extraction, extrapolation, and uncertainty. For ANN models to be more effective and useful in the future, it is essential that further work be undertaken in these four areas, particularly in the context of geotechnical engineering.