ارزيابي و پيش‌‌بيني فروپاشي خاك‌‌هاي رمبنده با توجه به مشخصه مكش آب در خاك با استفاده از يادگيري ماشين

The collapse potential of unsaturated soils presents a critical challenge in geotechnical engineering, particularly in arid an‎d semi-arid regions where matric suction an‎d inundation pressure significantly govern soil behavior. This study innovatively investigates the combined effects of matric suction, inundation pressure, an‎d san‎d content on the collapse potential of natural fine-grained soils, alongside conventional parameters such as dry unit weight, initial moisture content, an‎d fines content. Utilizing a modified hydraulic consolidation device capable of simultaneous controlled suction an‎d inundation pressure application, collapse potential was eva‎luated under suction levels ranging from 0 to 80 kPa an‎d inundation pressures from 100 to 400 kPa. Results revealed that matric suction can amplify collapse potential by up to 2.5 times due to increased effective stress an‎d the breakdown of capillary bonds upon saturation. Higher inundation pressures further intensified collapse severity, leading to classification shifts per ASTM D5333 stan‎dards. To enhance predictive capability, two empirical relationships are proposed: an empirical model predicting collapse risk based on matric suction, inundation pressure, an‎d san‎d content; an‎d a generalized equation for estimating collapse potential at various suctions from its zero-suction value, incorporating soil type effects. Furthermore, sensitivity analysis indicated that matric suction alone accounted for over 95% of the variability in collapse potential. In a parallel investigation, the influence of geological characteristics an‎d Soil–Water Characteristic Curve (SWCC) parameters on the collapse potential (CP) of unsaturated soils from loessic an‎d lacustrine–alluvial deposits in Iran was examined. Laboratory experiments, including SWCC determination, suction-controlled modified consolidation tests, an‎d microstructural analyses (SEM an‎d XRF), were conducted. The findings demonstrate that matric suction significantly affects compressibility an‎d collapse behavior, with higher suction an‎d Pressure at wetting increasing CP. SEM observations revealed microstructural rearrangements an‎d denser particle packing consistent with hydraulic an‎d mechanical responses. For improved prediction, machine learning models were trained on an integrated dataset, with Gradient Boosting achieving the highest predictive performance (R² = 0.859, RMSE = 1.273). Permutation importance analysis identified Pressure at wetting, matric suction, an‎d saturated volumetric water content (θs) as the most influential mechanistic predictors. These findings highlight a strong quantitative association between suction-controlled SWCC parameters an‎d collapse potential, indicating that the incorporation of hydraulic variables significantly improves predictive accuracy compared to conventional index-based approaches. The proposed integrated experimental–hydraulic–machine learning framework provides a reliable preliminary tool for collapse assessment an‎d strengthens the mechanistic basis for geotechnical design in unsaturated collapsible soils.

مهدي ابطحي فروشاني , حميد هاشم الحسيني

البرز حاجيان نيا

محمود قضاوي , ميثم مشايخي , محمدعلي روشن ضمير