A review of artificial neural networks used for estimating mechanical soil parameters
A review of artificial neural networks used for estimating mechanical soil parameters

Author(s): Mohammed Amin Benbouras, Ratiba Mitiche Kettab, Fatiha Debiche, Nassim Hallal, Maroua Lagaguine, Alexandru-Ionuţ Petrişor
Subject(s): Architecture
Published by: INCD URBAN-INCERC
Keywords: geotechnical parameters; in situ and laboratory tests; artificial intelligence techniques; artificial neural network
Summary/Abstract: In the geotechnical field it is possible to encounter very complex phenomena, which are hardly analyzed by analytical methods based on physical laws. Classical methods, such as empirical correlations have too little power to efficiently generate them. Nevertheless, geotechnical engineers have successfully adopted different experimental methods of identifying parameters based on in situ and laboratory tests. Several studies have employed new empirical mathematical approaches for treating and estimating soil parameters, in order to overcome the relatively expensive and time-consuming in situ and laboratory tests; they are called “artificial intelligence techniques”. This work presents a literature review of many applications aimed at estimating some soil parameters considered crucial to the decision-making process and important for the identification of geotechnical hazards. We focus on the artificial neural networks technique. This method is selected because it has been proved to be the most successful artificial intelligence approach applied to geotechnical engineering. As an example, two applications aimed at estimating the compressibility and swelling coefficients have been illustrated. The findings clearly indicate that ANN methods are of great importance for geotechnical engineering, as they have yielded cost-effective and valuable results for simulating the complex heterogeneous behavior of soil and efficiently estimated the geotechnical parameters.

• Details
• Contents