Artificial neural networks for predicting the hydraulic conductivity of coarse-grained soils


EURASIAN SOIL SCIENCE, vol.38, no.4, pp.392-398, 2005 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 38 Issue: 4
  • Publication Date: 2005
  • Title of Journal : EURASIAN SOIL SCIENCE
  • Page Numbers: pp.392-398


Artificial neural networks have been extensively used in soil mechanics to predict different soil behaviors. This paper is aimed at developing an ANN model to predict the hydraulic conductivity of granular soils to act in conjunction with field and laboratory tests. For use in the ANN model, 95 individual permeability tests, which were conducted on different heterogeneous porous granular media to calculate the hydraulic conductivity of a granular matrix with different soil compositions, were performed under laboratory conditions. Many of the test results (82 individual test results) were used to train the ANN system, and the rest of the experimental results (13 individual test results) were used to predict the hydraulic conductivity of the matrix. In order to train the ANN model, the fine-grained soil content, sand content, gravel content, grain size diameters d(10) and d(50) of the total soil mass, and void ratio were chosen as parameters affecting the hydraulic conductivity. Also, the ANN results were compared with those of the multiple linear regression method and two empirical formulas (Hazen and Slichter formulas), and it was seen that the ANN results were very encouraging. As a result of this research, an ANN model as a simple prediction tool, which calculates the permeability from different soil parameters, was developed.