Neural Network Modeling of the Mechanical Behavior of Sandby Glenn W. Ellis, Clarkson Univ, Potsdam, NY, USA,
Chengwan Yao, Clarkson Univ, Potsdam, NY, USA,
Rongda Zhao, Clarkson Univ, Potsdam, NY, USA,
Document Type: Proceeding Paper
Abstract: Given a set of triaxial test data, an artificial neural network (ANN) has been trained to model the mechanical behavior of a medium-to-fine sand. Once trained, the ANN was tested by simulating the triaxial test results for conditions not previously encountered by the network. In these simulations the ANN accurately modeled the influence of relative density and confining pressure on the mechanical behavior, including strain-softening and dilatancy characteristics. Poor learning resulting from overtraining the network was studied using artificially generated triaxial data. Absolute bias errors in the training set were found to be a cause of overtraining.
Subject Headings: Triaxial tests | Model tests | Neural networks | Mechanical properties | Mathematical models | Soil stress | Data processing
Services: Buy this book/Buy this article
Return to search