tensor grid manager

[qilicun]

Hi,
I'm trying to write an input file of SPE9 for testing opm-autodiff/sim_fibo_ad. I find that the tensor grid just support constant vector for the keyword "TOPS", but which in SPE9 is not constant yet, it's a nonuniform vector. So is there any other method to create a SPE9 grid or could the "TOPS" be enhanced to support nonuniform vector?
Thanks.

[atgeirr]

Enhancing TOPS to support non-constant top surface requires us to define just what we mean by it. For a simple example: given a grid of two cells with varying TOPS, should the grid become (1) a stair-step grid, (2) a conforming grid with sloping top/bottom faces, (3) or a stair-step grid where the connectivities are (magically) the same as for the conforming grid? Since we have not decided what it should mean, we do not support it. If you have suggestions as to the handling/meaning of TOPS, I am listening...

For now, I suggest using DEPTHZ (giving pillar top z values) or COORDS/ZCORN (full corner-point grid).

For the particular case of SPE9, we have made a version of that (very simple) grid using COORDS/ZCORN, this version makes the assumption that what you want is option (2) -- a conforming grid with parallelepiped cells. Please send me an email if this is what you want.

[qilicun]

@atgeirr Just as you said, enhancing TOPS is an undefined problems which I don't think much more about. What I want is trying to use TOPS create a conforming grid with sloping top/bottom faces especial for SPE9. DEPTHZ may be a good choice. By the way, could you send me your SPE9 corner grid file? And usually, which tools is used to create corner grid mesh file? Thanks.

[bska]

Just a small remark. All definitions I've ever seen of the SPE9 grid use the DX, DY, and DZ keywords in addition to TOPS. Strictly speaking, these keywords will (probably, I haven't tested it) allow specifying a geometry that cannot possibly exist in three space dimensions. The format does, in principle, allow specifying

DX(i) = i

and similarly for DY and DZ.

Therefore, whenever I have a choice, I personally prefer the more FrontSim-like keywords DXV, DYV, DZV and DEPTHZ. In the case of SPE9, these would be

DXV
  24*300
/

DYV
  25*300
/

DZV
  20 15 26 15 16 14   8
   8 18 12 19 18 20 50 100
/

and then the DEPTHZ would be the output of the following programme

#include <math.h>
#include <stddef.h>
#include <stdio.h>

int
main(void)
{
    size_t i, j;
    double x, dip;

    dip = tan(10.0 / 180.0); /* 10 degree dip angle */

    printf("DEPTHZ\n");
    for (j = 0; j < 26; j++) {
        for (i = 0; i < 25; i++) {
            x = 300.0 * i;
            printf(" %.10e", 9000.0 + x*dip);

            if ((i + 1) % 5 == 0) {
                printf("\n");
            }
        }
    }
    printf("/\n");

    return 0;
}

[bska]

dip = tan(10.0 / 180.0); /* 10 degree dip angle */

Well, almost. Of course, the real value would be something along the lines of

dip = tan(1.745329251994329e-01);  /* approximately pi/18 = 10*pi/180 */

Sorry about that.

[qilicun]

@bska Thank you for your comments, that's a another way for giving the grid file for SPE9, when you don't have the COORD/ZCORN mesh file. Thanks. :)

[bska]

There's been no activity on this issue for eight days and it appears to be resolved. I'll close the issue, but please reopen if the problem resurfaces.