A new nonlinear multivariable model for UCS prediction in underground multi-seam coal measures with high-plastic clay interbeds

OpenAlex İngilizce

Accurate estimation of uniaxial compressive strength (UCS) is critical for planning and optimising underground mining operations, yet direct UCS testing in weak, heterogeneous coal seams is costly and time-consuming. This study focuses on the development of nonlinear bivariate and multivariable models for predicting UCS of coal in high-plastic claystone interbedded thin coal seams. The distinctive aspect of the research is the comprehensive integration of physical, mechanical, dynamic elastic and geochemical data to predict UCS under complex geological conditions. A laboratory dataset containing 13 variables was used, including physical, mechanical, dynamic elastic and chemical properties of coal. Nonlinear exponential prediction models were established for two thin coal seams under challenging geological conditions, and the Levenberg–Marquardt algorithm was applied for model optimisation. Model performance was assessed through cross-validation and 1:1 line comparisons during verification and generalisation stages. The most accurate models incorporated porosity, density, Schmidt rebound hardness, Brazilian tensile strength, Hardgrove grindability index, impact strength index, compressional and shear wave velocities, ash, and carbon content. The optimised multivariable models achieved RMSE values of 0.21–0.25 MPa. The resulting UCS prediction framework offers a practical decision-support tool for mine design and operational planning in clay-interbedded underground coal measures.