Poison coefficient calibration

Hi,

This has been discussed in length in the DEM litterature since e.g. Micromechanical Aspects of Isotropic Granular Assemblies With Linear Contact Interactions, R. J. Bathurst and L. Rothenburg, Journal of Applied Mechanics, 1988.

Assuming you're using in YADE FrictMat contact models or derived classes, such as JCFpm model, the answer is "the ratio of normal over tangential contact stiffnesses", i.e. ........ poisson [*] !
(the name is completely misleading, though...)

[*] https://yade-dem.org/doc/yade.wrapper.html#yade.wrapper.JCFpmMat.poisson

(Assuming the strain discrepancies occur in the "elastic" regime, as your question title suggests..)

Hi Jérôme,

Thank you for the response.
And, yes, the strain discrepencies occur in the elastic regime.

Best regards.
Jabrane

The problem is that modifying this ratio will cause modification of other macromechanical properties such as the strength... it might be a non solvable problem in the end and this is actually a limitation of the model that you use (JCFPM), but you have to make some compromises at some point (like in any other modelling approaches).

Another thing: how is the lateral strain measured in your numerical experiment? Is it a local measurement like in real experiments (strain gauges) or is it through the displacement of the boundary walls? That would actually be interesting to compare these two scales to make sure you compare things that are comaprable.

Luc

Hi Luc,

In the experimental test the lateral strain is measured by the strain gauge put on the position h/2 (h: sample height) of the cylindric sample. In the link below, you find the video in which it is demonstrated how the uniaxial test is performed.
https://www.isrm.net/gca/index.php?id=1234

In the pure elastic phase the numerical lateral strain and the esperimental lateral strain are equal. But in the damage phase, they are not equal. In the link below, you find the stress/lateral strain (numerical and experimental) curves.

https://filex.univ-lorraine.fr/get?k=sy0sM5sPbTt7swz9sUb

To conclude, i try to minimize the discrepencies between the experimental latreal strain and the numerical lateral strain in the damage phase.

Best regards.
Jabrane.

The calibration of micromechanical parameters is not straightforward. As Luc mentions, each microparameter impacts multiple macroparameters. Most studies use some kind of iterative approach for calibration. However, I found one study that demonstrated a very clean and straightforward sensitivity analysis for calibration of deformational microparameters [1]. Strength parameters, on the other hand, are much more difficult to calibrate. In some cases people are even using probabilistic methods to find the best fitting parameters! [2]. At the end of the day, you need to make some decisions on what kind of error you are willing to accept, and where you are willing to accept that error given the differences between the simulation and experimental setup.

[1] Wang, Y., & Tonon, F. (2010). Calibration of a discrete element model for intact rock up to its peak strength. International journal for numerical and analytical methods in geomechanics, 34(5), 447-469.

[2] Zhang, Y. B., Medina-Cedina, Z., & Khoa, H. D. V. (2011). Probabilistic Calibration of a Discrete Particle Model for Geomaterials. In Geo-Frontiers 2011: Advances in Geotechnical Engineering (pp. 4204-4213).