Servo wall - high target values lead to a lost of particles

Submitted by Alaska on Mon, 03/16/2015 - 11:04

Hi,

I am developing a numerical ring shear test. For that I am using a rectangular 3D box with a size of 8e-2 * 8e-2 * 4.4e-2 m. I fill this box with ~25.000 particles with sizes of 9.2e-4 to 16.4e-4 m. After settling under gravity I want to compress the probe with a servo wall and a target value of -5e6 Pa (in z direction).

My particles have follwing parameters, following Aigner et al., 2013 "Determing the coefficient of friction by shear tester simulation":

youngsModulus = 65.0e9
poissonsRatio = 0.45
coefficientRestitution = 0.9
coefficientFriction = 0.6
coefficientRollingFriction = 0.06

When I start the compression:

fix cad 1 all mesh/surface/stress/servo file top_shear.stl type 1 scale 0.9999 stress on com 4e-2 4e-2 2.2e-2 ctrlPV force axis 0 0 1 vel_max 0.01 target_val -5e6

unfortunately, the particles disappear the model after some compression steps. When I tried to compress the probe using a lower target value a compression of around 5000 Pa is possible and the particles stay in the model. Higher values lead to the phenomenon that the particles get lost (and disappear from the model).

Do you have any suggestions with that problem?

Thanks in advance.

Forums:

ckloss's picture

ckloss | Mon, 03/16/2015 - 12:17

Hi Alaska,

>>Do you have any suggestions with that problem?
I am not really experienced with using the servo wall, but it is somehow sensitive to the pairing of particle properties and controller parameters
Also, are you using the latest version of LIGGGHTS(R)?

Christoph

Alaska | Mon, 03/16/2015 - 14:04

Dear Christoph,

thanks for your quick reply. I am using the parameters for the particles you used in your paper (mentioned above, at least I believe it is yours :) ). Unfortunately, I can not find the target value and maximum velocity of the servo wall you used for the compression in your model.

>>Also, are you using the latest version of LIGGGHTS(R)?
Yes, I am using the latest version of LIGGGHTS(R) - Version 3.1.

Do you have any suggestions how to pair my particle properties and the controller parameters to avoid disappearing particles? What exactly are the controller parameters - the ones of the servo mechanism?
And what are the factors causing a disappear of particles?

If you have a time for a look in my scripts I would like to send them via PM.

Thank you so much.

Cheers, Alaska

JoshuaP | Mon, 03/16/2015 - 14:19

you know that the target_value is not in Pa? It is not a pressure, it is a force. So you have to calculate your target_force by pressure*area.

regards

Alaska | Mon, 03/16/2015 - 14:34

Hi Joshua,

thank you so much for your reply. I did not know that. I just wanted to double check: my area is 8e-2 * 8e-2 m and I multiply that by 5e6 (my target value) which is given in Pa, right? Means I have a target value of 32.000. Which is still high. And my particles tend to disappear again.
Do have any further suggestions? Or did I missunderstand you?

Best, Alaska

JoshuaP | Mon, 03/16/2015 - 14:41

thats right. Target value should be 32000 for a pressure of 5 MPa. Maybe drive slower or reduce the timestep.

Alaska | Mon, 03/16/2015 - 14:49

Thanks again :) I will try it with a lower velocity (that will take some time). My timestep is with 0.000001 already quite small.

I will keep you updated!

JoshuaP | Mon, 03/16/2015 - 14:33

btw you could try to reduce the velocity of the servowall or to reduce the timestep.
Maybe post the necessary parts of your script for the servowall?

Alaska | Mon, 03/16/2015 - 14:57

Thank you so much for all your answers and input. I will try to set down the rolling friction if the lower velocity of the wall is not working ( I am actually testing that). I am presently using the hertz tangential pair style.

I will keep you in the loop :)

Cheers, Alaska

Alaska | Tue, 03/17/2015 - 14:08

Dear All,

thanks again for all your answers. I tried to compress my probe with really low velocity and without rolling friction. Unfortunately the particles still disappear :( Do you have any further suggestions what this could cause?

Best, Alaska

JoshuaP | Tue, 03/17/2015 - 15:52

you have very stiff particles, maybe try a smaller time step :
variable deltatime equal sqrt(${rho}/($E/(2*(1+$v))))*c_minradi*PI/(0.1631*$v+0.876605)/10
timestep ${deltatime}
where rho is particle density E is youngsmodulus c_minradi is the smallest radius used and v is poisson number

Alaska | Wed, 03/18/2015 - 14:08

Hi Joshua,

I checked my timestep and it is very, very small. Your equation gave me a timestep of 0.0001 and my one is 0.000001.

However, I solved the problem. The Young´s Modulus was too high. With a lower value of 65.0e8 (instead of 65.0e9) I can compress my particles with 5e6 Pa (means a force of 32000N) and all particles stay in the model.

Do you have suggestions which value I have to use for modelling a quartz sand?

Thanks you all again fpr your help and suggestions!
Cheers,
Alaska

JoshuaP | Wed, 03/18/2015 - 15:24

You decreased the youngs modulus, so what happened is that now your timestep was small enough. I think for the higher youngs modulus you need just a smaller timestep. I didnt investigated the youngs modulus that much, but I think it depends also on what you are modeling. For a high compression simulation the youngs modulus becomes important and I would try to compare compression results in DEM with real experiments.
Maybe look for oedometer test.

regards