Evaluation method for collapsible deformation of thick loess sites

• First identified a 21–24 m collapse transition zone in thick loess via real-time stress monitoring. • Proposed a dual-coefficient model for accurate collapsibility evaluation of thick loess sites. • Revised β₀ from 1.5 to 1.9 in Chinese Code GB 50,025–2018 for shallow loess to prevent collapsibility underestimation. • A collapse probability coefficient α = 0.4 is introduced for deep loess (>24 m) to prevent overestimation. • Achieved high collapsibility prediction accuracy with the new framework. The rapid urbanization in the collapsible loess regions of Northwest China has heightened the critical need for accurate evaluation of collapsibility in thick loess sites (≥15 m). This study investigates the self-weight collapsible deformation characteristics of thick loess through a large-scale field immersion test (test pit diameter: 24 m, total water injection volume: 13,000 m³, monitoring duration: 163 days) and the analysis of 18 sets of engineering data from the Longxi region. Based on real-time in-situ stress monitoring—a novel approach in such tests—this study, for the first time, identifies a distinct transitional zone at 21–24 m depth, which exhibits an abrupt change in collapsibility and forms the interface between complete and incomplete collapse. Below 24 m, the collapsibility potential of the soil is significantly constrained, with measured vertical stresses reaching only 11 %-35 % of the theoretical saturated overburden stress. To address the limitations of the current code (GB 50,025–2018), which employs a single correction factor and fails to capture this depth-dependent behavior, we propose a pioneering dual-coefficient correction framework. For loess sites with thickness < 24 m in the Longxi region, the current regional correction coefficient β 0 (1.5) substantially underestimates collapsibility and is critically revised upward to 1.9 (linear regression slope: 1.89, R² = 0.98). More innovatively, for deep collapsible loess layers (> 24 m), we introduce a novel collapse probability coefficient α (recommended value: 0.4) to replace conventional correction coefficients, thereby preventing the overestimation of collapsibility. This research provides a refined and more reliable theoretical and experimental foundation for precise collapsibility evaluation in deep loess regions, offering practical guidance for similar engineering projects.