MATHEMATICAL-STATISTICAL AND COMPUTER MODELING METHODS USED IN ESTIMATING AND PREDICTING DEFORMATION INDICATORS

Abstract

In the article Fergana in the province wide widespread loess-like, alluvial-proluvial and irrigated small particulate of the ground variable voltages in the conditions deformation behavior assessment and prophecy in doing applicable mathematical-statistical and computer modeling methods scientific in terms of is based on. In the study laboratory ( cyclic triaxial, consolidation, collapse ) and field ( plate) loading, LWD and others ) tests from the results consists of information base formed and based on it resilient module M_r, permanent deformation ε_p, collapse potential I_cand settlement indicators are determined, their statistical structure, distribution and correlation relationships. Using multifactor regression models, simple and physically understandable equations are developed to estimate M_rand I_c, and their accuracy is analyzed in the cross-section of geotechnical zones. In order to take into account nonlinear relationships in more depth, predictive models based on artificial neural networks (ANN) are built, providing a significant improvement in R² and RMSE indicators compared to regression models. At the same time, settlement and deformations in the foundation-soil system are estimated using computer models based on the finite element method (FEM). territorial and depth according to distribution, moisture and collapse under the influence differential sediments dynamics is studied and laboratory-field results with verification is done. Obtained results based on regression, ANN and FEM approaches integrating regional deformation prediction methodology offer in the conditions of the Fergana region buildings the foundation reliable design, deformation danger zones separation and geotechnical risks reduce for scientific and methodological basis created.