Triaxial compression test

Hi all,

I send you a first idea to control the wall velocity by keeping the stress constant. If we manage to do it, then we'll be able to implement the "numerical membrane" to perform a triaxial test on a cylindrical shaped assembly of particles.

I wrote these few lines from the fix_move_tri.cpp file. It is not a perfect script, it is just a quick idea !!!

I think that we will need to add this possibility in the script:

fix ID group-ID all move/mesh/gran stress_applied X Y Z units box wall-type

In the function void FixMoveTri::initial_integrate(int vflag) inside the fix_move_tric.cpp file:

void FixMoveTri::initial_integrate(int vflag)
{

double max_vel = 0.001; # by default
double depth = yhi-ylo;
double height = zhi-zlo;
double width = xhi-xlo;
double force_applied = 0;
double translation = 0;

double *ftrans;
force_trans = (double *) calloc ((size_t) 3, sizeof (double));

double max_vel1 = max_vel * 3 * width /(depth+height+width);
double max_vel2 = max_vel * 3 * depth /(depth+height+width);
double max_vel3 = max_vel * 3 * height /(depth+height+width);

double delta = (update->ntimestep - time_origin) * dt;

if (mstyle == LINEAR) {
for (int i = 0; i < nlocal; i++) {

if (vxflag) {
force_applied = stress_applied[0]* Surface_triangle;
force_trans[0] = stress_applied[0] - f[i][0]; #if f[i][0] is the resultant force due to the spheres
translation = force_trans[0]/ normal_wall_stifness;

x[i][0] += std::min(std::abs(translation), delta * max_vel1) * Mathr::Sign(translation);
}

}

.....

You have the main idea that I propose, I need to read more the script to know the best way to implement these lines.
It remains few developments to do.

See you soon
Jf

Hi george,

what you propose sounds interesting, but I have several questions :-)

>>I am not sure how appropriate it is to have a shrinking but rigid cylindrical wall.
Can you explain this (I am not an expert in triaxial stress tests) - making a shrinkable cylindrial wall was my first guess. I agree with you that you might get in trouble if the deformation is too large - but if it is just like 1%...?

>>code that would check for the grains on the outside of your sample, triangulate the surface and apply a force on each >>proportional to the area that each maps to
Which surface would you need to triangulate? What do you mean with "area that each maps to"?

How are these tests conducted in reality - I think there a movable plates that keep stress constant right ?

I will be happy to discuss these issues with you!

Christoph