Triaxial compression behavior of sand and gravel using artificial neural networks (ANN)

The stress-strain and volume change behavior of sand and gravel under drain ed triaxial compression test conditions was modeled using feed-back artific ial neural networks. A large experimental database obtained from published literature was used in training, testing, and prediction phases of three ne ural network based soil models. Issues related to the number of hidden unit s, magnitude of strain increment during feed-back, and over-training error are discussed. These models can accurately represent the effects of mineral ogy, grain shape and size distribution, void ratio, and confining pressure. The observed behavior in terms of a nonlinear stress-strain relation, comp ressive volume change at low stress levels, and volume expansion at high st ress levels are captured well by these models. (C) 1999 Published by Elsevi er Science Ltd. All rights reserved.