DESCRIPTION OF THE ENGINEERING GEOLOGICAL CONDITIONS OF THE FERGANA REGION AND WIDELY DISTRIBUTED SOIL TYPES

Abstract

The article proposes a “regional deformation prediction methodology based on cyclic loading and artificial intelligence” for predicting changes in some deformation parameters of soils under variable stresses. Loess, alluvial-proluvial and fine-grained soils widespread in the Fergana region are selected as the research object, and their physical and mechanical properties, wetting regime and deformation behavior under the influence of seismic and transport loads are comprehensively studied. First, laboratory tests are conducted under cyclic loading, and an experimental database on resilient modulus and permanent deformations is formed. Then, the parameters of the Cam-Clay type constitutive models are recalibrated based on regional soil properties, and the results obtained are integrated with a prediction model developed using artificial neural networks and other ML approaches. The proposed methodology allows for a high-accuracy prediction of deformation parameters, taking into account the effects of soil collapse potential, moisture and stress levels. This approach, adapted to the engineering-geological conditions of the Fergana region, is of practical importance for designing the foundations of residential and industrial structures, assessing the long-term stability of road and irrigation structures, and reducing regional geotechnical risks.