Experimental study for modeling the unloading swelling behavior of Ariake clay

Due to economic and demographic growth, there is a rising demand for land reclamation in coastal cities of East and Southeast Asia. Marine clays typically play a critical role in these projects, and the deformation characteristics of marine clays become a crucial problem in terms of the quality of the subsoil conditions. The long-term loading behavior of marine clays has been studied by many researchers. However, relatively few studies have been done on the unloading behavior of these clays after preloading; and thus, the strain rate dependency on the unloading behavior of marine clays remains unclear. The aim of this study was to accumulate experimental data on the unloading behavior of marine clays and to develop a strain rate-based model for improving the accuracy of the predictions of the swelling behavior of marine clays during unloading. The authors conducted a series of constant rate of strain (CRS) consolidation tests from loading to unloading, and long-term unloading oedometer tests on Ariake clay, which is a well-known sensitive marine clay, to observe the swelling behavior during in unloading. The preloading time, corresponding to different strain rates at the end of preloading, was controlled to elucidate the effect of the stress history. Moreover, instead of parameter σ′ p (preconsolidation pressure) for the normal consolidation visco-plastic behavior, the authors developed and proposed a new visco-plastic model by introducing the concept of a plastic rebound boundary and a new parameter R for swelling behavior during unloading. Parameter R represents the normalized distance from the current stress state to the plastic rebound boundary in logarithmic effective consolidation stress. Therefore, the visco-plastic model for the behavior in the loading stage was developed into the swelling visco-plastic behavior in the unloading stage for Ariake clay. Comparing the simulation and test results, the simplified visco-plastic swelling model was found to agree well with the test results.