Introduction. The paper compares the effectiveness of different approximation and interpolation methods for reconstructing soil property values from static sensing results in plan and depth.

Aim. The aim of the study was to evaluate the effectiveness and develop a methodology for building geostatistical models based on kriging to recover values of soil properties and to predict non-uniform settlement of foundations of high-rise buildings and structures.

Materials and methods. Uneven foundation settlement was determined on the basis of the recovered values of soil properties, taking into account the statistical heterogeneity of the soil mass using the finite element method (FEM) in the Plaxis 3D software package. Special attention was paid to geostatistical methods based on kriging, which allow a deeper analysis of geomechanical data.

Results.

1. Comparative analysis of different approximation and interpolation methods has shown that geostatistical models based on kriging, provide high accuracy when recovering soil property values at unknown points.

2. The developed methodology allowed us to build a geostatistical model of heterogeneity of soil massifs based on the results of geotechnical engineering surveys.

3. Numerical modeling in Plaxis PC confirmed the effectiveness of the proposed approach, showing that the computational model with heterogeneous soil massif built with the help of kriging accurately reproduces the real conditions than the model with constant soil parameters.

Conclusions. The use of geostatistical models based on kriging increases the accuracy of estimating the values of building settlement and roll, taking into account the heterogeneity of the soil mass.

The proposed methodology of building a geostatistical model based on kriging allows to estimate effectively the properties of the soil in plan and depth.

The results of numerical simulation of the computational model have shown that the use of spatial interpolation based on kriging to determine the values of the soil massif parameter gives results close to the data obtained with the full set of initial data.