Edinburgh contact stiffnesses

Submitted by kees_bulk on Fri, 02/25/2022 - 13:33

Hi,

I am working with the EEPA/Edinburgh contact model and due to the lack of doc something is unclear to me.

There are two input stiffnesses called CoeffAdhesionStiffness and UnloadingStiffness. I was unable to exactly determine what the value of the CoeffAdhesionStiffness (kn2kc) should be as it is not used the same way in the original work by Morrissey. In the code it is used to multiply with the loading stiffness to get the adhesion stiffness:

k_adh = kn2kc[itype][jtype] * kn

So should it be a relatively small number then as it is probably just a coefficient?

Regarding the UnloadingStiffness (kn2k1), in the code it is used in a similar fashion by multiplying with the loading stiffness:

k2 = kn*kn2k1[itype][jtype]

But if I put a small coefficient (e.g. 1.5) in, the forces are not transferred between particles and thus it is too low. Should the input value then simply be the actual UnloadingStiffness?

Forums:

Nathan | Tue, 03/15/2022 - 11:36

Hi,

Are you developing your own code for EEPA model or you employed an existing one in LIGGGHTS? You can look at some examples of this contact model developed in PFC, quite lots of information.

Regards,

kees_bulk | Tue, 03/15/2022 - 14:06

I was trying to work with the LIGGGHTS built-in "normal_model_edinburgh.h". After thorough examination I got an answer to my stiffnesses question by debugging it, namely that the UnloadingStiffness is the k1/k2 ratio and the CoefficientAdhesionStiffness portion in the code is not used as it is outside an if-loop that is permanently true.

However, to me it still seems incorrect as it uses terms from the linear model when using the non-linear mode.

I have indeed looked at the PFC EEPA model as it is better documented, but the different approach of both programs makes it hard to copy parts of it. I am now editing the LIGGGHTS model to what it should be according to me from literature to see whether its performance will be closer to what I expect, although the downside is that I do not have enough time to perform extensive verification.

Mainly I was curious whether others have succesfully used this model or that they used different approaches.

Nathan | Wed, 03/16/2022 - 13:37

Hi,

I found some discussions in the forum that someone used the models but none of them verified or reported their performance. According to LIGGGHTS releases 2018, the model was developed by Rahul Mohanty (P&G and University of Edinburgh) and Tomaz Zorec (University of Ljubljana). I thought some must has double checked the code before incorporating into LIGGGHTS. But when I searched this code in the latest version of LIGGGHTS, I could not find it well. Was it removed from the Github of LIGGGHTS? Have you compared the model results by LIGGGHTS and PFC? Need lots of efforts to edit and verify the code in my point of view. We can work together if possible.

Regards,

kees_bulk | Wed, 03/23/2022 - 09:14

Hi,

I also assumed some must have verified it before officially adding it to the source code, but with the lack of documentation or reports it is hard to validate. It does exist in the official 3.8.0 github version of LIGGGHTS under the name 'normal_model_edinburgh.h'.

I currently found out that my previously wrong results were due to not turning on the tangential history for the particles and now it gives comparable results compared to literature. (I only tried an angle of repose test yet). I have edited some parts which seemed incorrect to me is adding the tangential multiplier to the tangential stiffness k_t and the nonlinear terms to the damping coefficients if the exponent n is higher than 1. Still hard to thouroughly verify whether it is correct, but it will make do for now.

Nathan | Fri, 03/25/2022 - 00:52

Hi Kees,

That sounds great that you achieved reasonable results, at least for the angle of repose. I will do myself the validation too. We have some laboratory tests with cohesive materials and see how it goes. If possible, can I have your contact email, we can exchange some results to improve our experience and understanding.

Thanks,

Regards,

kees_bulk | Wed, 04/20/2022 - 09:53

Hi Nathan,

Sorry for my late response, I haven't checked this thread in a while. How is your validation going?

If you could provide your contact email I will contact you.