Authors
Abstract
Artificial Neural Networks (ANNs) are used to relate the properties of gypseous soils
and evaluate the values of compression of soils under different conditions. Therefore, onelayer
perception training using back propagation algorithm is used to assess the validity of
application of ANNs for modelling the settlement ratio for wetting process, (S/B)w, and the
settlement ratio for soaking process, (S/B)s.
It was found that ANNs have the ability to predict the compression of gypseous soil
due to soaking, washing process with high degree of accuracy. Also, performance of ANNs
showed that one hidden layer with one hidden nodes is practically enough for the neural
network analysis.
The sensitivity analysis indicates that the viscosity and specific gravity have the
most significant effect on the predicated settlement ratio and the density of injection material
and void ratio have moderate impact on the settlement ratio. The results also show that the
initial gypsum content, stress and time have the smallest impact on settlement ratio.
It was concluded that the artificial neural networks (ANNs) have the ability to
predict the settlement ratio for wetting process (S/B)w, and settlement ratio for soaking
process (S/B)s of gypseous soil with high degree of accuracy. The equations obtained using
(ANNs) for (S/B)w, and (S/B)s showed excellent correlation with experimental results where
the coefficients of correlation are (0.9541) and (0.991), respectively.
Keywords
)
