Dear All,
I am using sand hypoplastic model in my research and I have encountered some problem when looking into cyclic simple shear test in the range of small strains.
Firstly, I have tried to simulate cyclic simple shear test from Shahnazari & Towhata (2002), Fig. 5, on Toyoura Sand. It is in the range of medium strains, up to 3%. The simulation looks really good (see attached pdf file). The sample contracts systematically with cycles thus the stiffness increases as per the experiment. Note contraction is with negative sign, Abaqus standard notation.
Secondly, I have subjected the same sample to cyclic loading but with significantly reduced strain amplitude, i.e. 0.1%. As expected, also here the sample contracts, however, the stiffness decreases with cycles which to me does not seem physical.
The model is calibrated as shown in Herle & Gudehus (1999) Table VII for Toyoura Sand. Intergranular strain parameters are set to also rather common values: R=0.0001, m_R=5.0, m_T=2.0, Beta=0.5, ksi=6.0. I have tried using both umats, from prof. Masin and prof. Fellin, both give the same answers.
I am wondering where is it I can be wrong here. Theoretically, I think I should be able to model sand behaviour for both strain levels with a single set of calibrated parameters.
Thank you all in advance for any comments and tips on how to approach my problem.
Regards
Dear Piotr,
The simulation in the range of medium strains looks really good and the parameters are typical, so I consider that the simulations are well performed. However, the behavior under small strains is unrealistic. In my opinion, this is a model limitation. Some authors have demonstrate the difficulty of reproducing different kinds of test conditions with a single set of parameters (which is a well-known drawback of the models, and is discussed in some works e.g. [1])
Regards,
Jose Duque
[1] Wichtmann, T. (2016), “Soil behaviour under cyclic loading: Experimental observations, constitutive
description and applications”, Habilitation, karlsruhe institute of technology (KIT).
Hey Jose
Thanks for the reply. It also seems to me to be a case as you described.
Actually, using more recent calibration from prof. Herle, with quite different set of the model parameters, seems to improve a bit the problem, not completely though.
Regards