مدل سازي شاخص فشردگي خاك هاي ريزدانه به كمك شبكه‌ي عصبي مصنوعي و مقايسه با ساير روابط تجربي

Modeling the Compression Index of Fine Soils Using Artificial Neural Network and Comparison with the other Empirical Equations

Construction of buildings and different structures leads to soil consolidation and as a result to soil settlement. Soil settlement is a function of variety of factors such as pressure deformation, depletion of pore water and etc. One way for estimating the soil settlement is to use the compression index which can be determined through consolidation test. Determination of this index in laboratory is time consuming. Therefore, in recent decades the researches have tried to relate this coefficient to some soil parameters, such as plastic limit, liquid limit, void ratio, specific gravity and so on, which can be easily measured in laboratory. There are therefore many empirical equations in the literature in this regard. In this paper the correlation of fine soil properties and compression index has been investigated using artificial neural network (ANN). A comparison was also carried out between the measured compression index in laboratory with the corresponding values obtained from the empirical equations and ANN model. The results showed that the Rendon-Herrero relationship calculates this index much better than the other considered empirical equations with the highest correlation coefficient and minimum error. It was found that the ANN model performed better than the Rendon-Herrero formula with higher accuracy in estimating the compression index. It was also found that the calibration of the coefficients in Rendon-Herrero formula from the existing data does not significantly improve the accuracy of this equation. Key words: Artificial Neural Network, Fine soils, Physical characteristics of soil, Soil compressing index