estimating the sand shear strength from its grain characteristics using an artificial neural network model and multiple regression analysis

Determination of soil shear strength is always among the most important issues in geotechnical problems. in this research, various neural network models and multiple regression are developed to obtain shear strength parameter of the sandy soil from physical parameters of roundness (r), maximum and minimum dry densities (γdmax, γdmin), relative density (dr), and grain sizes, d10, d30, d50, and d60. firstly, the effect of these physical parameters on the shear strength of sands is examined by soil laboratory tests. for this purpose, laboratory tests of the direct shear, maximum and minimum dry densities, and sieve analysis are conducted. subsequently, the laboratory results are used as a data set to develop an artificial neural network and multiple regression models to predict shear strength parameters. finally, the efficiency and appropriateness of each approach are discussed. results showed that both neural network and regression are precise, appropriate, and inexpensive methods to predict soil shear strength parameters.