پيش بيني مقاومت برشي درزه هاي سنگي پرشده با استفاده از شبكه هاي عصبي مصنوعي و مقايسه با مدل هاي تجربي

PREDICTION OF THE SHEAR STRENGTH OF INFILLED ROCK JOINTS USING ARTIFICIAL NEURAL NETWORKS AND COMPARISON WITH EMPIRICAL MODELS

Determination of the mechanical parameters of rock joints and mainly the influence of infill material of joints on the shear strength of discontinuous media has a great importance. For achieving this goal, till many researches have conducted on the simulated models with the use of physical modeling and laboratory testing. However, many of them are rather unemployable for some reasons such as different conditions of the surface and underground environments. Accordingly, in this paper an exact model has been developed for prediction of the shear strength of in filled rock joints using surface sampling of natural joints, numerous direct shear tests in different filling conditions and finally by utilization of the artificial neural networks. In order to increase the network accuracy, kinds of filling materials separated and a network with two hidden layers has been developed for each kind. The comparison of the results with empirical models demonstrated that these networks have better capability in prediction of the parameter in question