Staged Wetting and Contact Stress Redistribution Beneath a Segmented Hydraulic Structure on Collapsible Loess

Wetting-induced collapse of loess can produce large differential movements beneath irrigation structures even under modest nominal foundation pressure. This study investigates the coupled response of a segmented reinforced-concrete chute and its loess-based model foundation in a single 1:50 physical model explicitly interpreted as a 1 g kinematic mechanism test, with simultaneous wetting from the upstream and downstream pools. At corresponding scaled depths, the model-to-prototype geostatic stress ratio is (γm/γp)/50, approximately 1/50 for comparable soil unit weights; applied contact loading does not restore the prototype self-weight stress profile. Vertical movement records from eight markers, including an incomplete early Zh record, were evaluated together with contact and embedded pressure-cell measurements. In this test, contact stress at a softened support decreased from 154 to 30 kPa (80.5% unloading), while stress at the adjacent initially dry support increased from 150 to 220 kPa (46.7% amplification). The flume-end stress decrease of approximately 39% is interpreted as being consistent with beam action on a shrinking temporary central support. This support configuration is a proposed explanation of the observations; its pressure and contact area were not measured. Final settlements at the eight markers had a spatial coefficient of variation of 40.2%, describing within-test settlement nonuniformity rather than repeat-test variability or measurement uncertainty. Model settlements of 1.8 and 2.0 cm have geometric equivalents of 0.90 and 1.00 m at a scale factor of 50. Compared with field settlements of 0.76 and 0.93 m, the absolute differences are 0.14 and 0.07 m, respectively, corresponding to point-wise absolute percentage differences of 18.42% and 7.53%. These two comparisons are descriptive and do not validate prototype predictions. The indicator-grid deformation suggests a localized punching pattern, but one with unquantified boundary-location uncertainty. All quantitative findings are specific to this single test. The observed stress redistribution motivates a qualitative staged assessment framework; replication at different initial densities and water contents and independent verification are required before quantitative design application.