R Secci

Artificial Neural Networks and Kriging Method for Slope Geomechanical Characterization

In this work, Multi Layer Perceptron Artificial Neural Networks (MLP ANNs) and Kriging method are applied for slope stability analysis. Both methods have been applied in order to evaluate detrital layer depth within a test site located in countryside of Capoterra (South Sardinia, Italy). Test site consists in a large area subjected to flooding and great magnitude debris flow events. Identified area stability and strength have been analysed by building a local geodatabase that allowed to perform a correlation analysis between depth of the detrital layers and respective geotechnical, geo-mechanical, hydraulic characteristics. Some other features regarding morphological, geological, structural, physiographic and vegetation settings have been considered. The comparison between the results obtained with the MLP ANNs and kriging method shows that the two methods can be applied to implement a realistic and accurate representation of the depth and geomechanical properties of incoherent deposits.