Abstract:

A series of dynamic triaxial tests were conducted on saturated remolded loess specimens under various loading conditions. A method of using hysteretic loops for each loading step to construct the backbone curve is developed. This method can utilize most of the testing information and reduce the randomicity of the tests compared with the traditional method of using only one hysteretic loop for each loading step. Hardin-Drnevich model, modified Hardin-Drnevich model and Martin-Darvidenkov model were then used to simulate the test results. Using this method, the accuracy and parameter sensibility of the above three backbone curve models were discussed respectively. Results show that Hardin-Drnevich Model can’t always simulate test data accurately in different test conditions. Modified Hardin-Drnevich Model and Martin-Darvidenkov model are more effective to simulate test data and they can reflect the softening characteristics of the backbone curve. Considering that Martin-Darvidenkov model is more sensitive than modified Hardin-Drnevich model, modified Hardin-Drnevich model is suggested to be used in the dynamic simulation of loess under complex loads.

Key words: loess, dynamical triaxial test, accumulated residual strain, backbone curve