一种黄土湿陷性原位评价技术

An in-situ evaluation technique for collapsible loess was developed using in-situ measurements and mathematic models characterizing collapsibility. The volumetric water content and natural density of loess were measured by time-domain reflectometry and simple cutting ring method, respectively. The collapsibility was evaluated using volumetric water content, natural density and physical properties through the mathematic models. The technique was verified based on 200 tests in five sites. Comparing with standard methods, more than 95% measurements of volumetric water content and natural density have an absolute error of ±3% and a relative error of ±2%, respectively. On the other hand, the models based on machine learning and elasto-plastic theory provide 88% and 83% accuracies in evaluating collapsibility (i. e. no, light, medium, and strong), respectively. The absolute errors of 90% calculated collapse potentials are within ±0. 01, ±0. 03 and ±0. 06 at light, medium and strong collapse, respectively.