What is the impact of having tetrahedra in a FLAC3D mesh

FLAC3D
1

Hi,

I am having a challenge with a mesh in FLAC3D 9.4 I generated in Griddle 3.0.

I am solving a creep problem with a relatively complex geometry which led Griddle to generate quiet a lot of pyramid (~14% of zones) and tetrahedra (~10%) and it seems to have on impact on the quality of my solution. As you can see below, the curvature of the cavern edge and the flatness of the plans above force Griddle to generate a transition zone with a lot of pyramid and tetrahedra.

If I look at the minimal principal stress with volumetric averaging contour the results seems fine as seen below:

image
image686×571 98.4 KB

But if I uses the constant contour I get these small single zone high and low stresse:

image
image699×621 99.4 KB

When I explore a bit further, by only showing zones with a high deviation in stress from there surrounding ( I a python script and save a flag to extra to be able to generate a threshold) I get the plot below: What is clear is that it does not correlate to zone condition, but seems to nearly only appear in pyramid and tetrahedra zones.

image
image743×550 101 KB

I seems that having tetrahedra in you mesh is much worst for the solution quality than I expected.

My questions are:

  • Have other experience with this issue? And what has been your solution.
  • Are there tips to solve it or the solution is just to get better at Griddle and generate a mesh with less tetrahedra and pyramids?
  • Does having tetrahedra is that bad for the numerical solution, or there something wrong with my simulation and it should not give these type of results?

Below is extract of GVol

*********** Histogram QS for All Elements ***********
Total number of bins : 11
Total number of counts : 2469469
Number of larger values : 0
Number of smaller values : 0
V max : 9.972195E-01
V mean : 7.948911E-01
V min : 1.400534E-02

Bin number – Bin boundaries – Hits

10 0.90 1.00 1082314
9 0.80 0.90 365978
8 0.70 0.80 292092
7 0.60 0.70 290576
6 0.50 0.60 214070
5 0.40 0.50 130941
4 0.30 0.40 71136
3 0.20 0.30 21318
2 0.10 0.20 1027
1 0.01 0.10 17
0 0.00 0.01 0

To be clear it does not seem to be an issue of zone quality, as other mesh I created previously Had better QS (V min above 0.2) and they had a larger problem with anomalous stress in zones.

Thank you in advance.

2

To add some clarifications, this a salt creep problem in which I apply a stress boundary condition on the wall of the cavern.

Also the issue seems to only appear when I solve the creep problem. If I solve the problem as a pure mechanical problem (assuming the material is elastic) the issue does not appear.

3

The tetrahedral elements in FLAC3D are overconstrained and prone to hourglassing effects, somewhat limiting their usability for creep and plastic deformation. To a lesser extent this also happens in wedges/pyramids sometimes. That is a known problem and the reason why users prefer brick elements (which effectively are also multiple tetrahedra, but their averaging procedure eliminates the problems).

To address this, the command zone nodal-mixed-discretization on may be used, introducing a similar averaging procedure for single tetrahedral elements based on the average values at their gridpoints.

Other than that, there is not much you can do, except creating an all-hex mesh, which is often not feasible in more “realistic” geometries. For the cavern in your images that should be possible, though.