sandbox/jieyun/test/vortex_ebit.c

    Single Vortex with the EBIT method

    The single vortex test with a highly streched and deformed interface was proposed by Rider, 1998. A divergence-free velocity field (u, v) = (\partial \phi \big/ \partial y, -\partial \phi \big/ \partial x) described by the stream function \phi = \pi^{-1} \sin^2(\pi x) \sin^2(\pi y) \cos(\pi t / T) is imposed.

    #define ADAPT 1

    The volume fraction (f) is only used for statistic purpose.

    scalar f[];
scalar * interfaces = {f}, * tracers = NULL;

#include "advection-ebit.h"
event stability (i++) {} // this ensures that time step is computed before marker advection
#include "ebit-2d.h"

const char *OUTNAME = "vortex";
const double xcenter = 0.5, ycenter = 0.75, R = 0.15;
const double Cfl = 0.125 [0];
double EndT = 2. [0, 1];
double ddt;
int IT, level;

int main() {
  EndT = 2.;
  for (level = 7; level < 8; level++) {
    init_grid (1 << level);
    ddt = 1. [0, 1]*Cfl/N;
    IT = (int) (EndT*N/Cfl);
    run();
  }

  EndT = 8.;
  for (level = 7; level < 8; level++) {
    init_grid (1 << level);
    ddt = 1. [0, 1]*Cfl/N;
    IT = (int) (EndT*N/Cfl);
    run();
  }
}

event init (i = 0) {
  vertex scalar phi[];
  foreach_vertex()
    phi[] = sq(R) - (sq(x - xcenter) + sq(y - ycenter));

  init_markers (phi);
  init_circle (xcenter, ycenter, R, f, s);

  semu2vof();
  area0 = area;
}

    The timestep dt and the velocity field are set.

    event stability (i++, i < IT, first) {
  dt = dtnext (ddt);

  double cdt = 1. [1, -1]*cos(pi*tTime/EndT);
  foreach() {
    double x0 = x/L0, y0 = y/L0;
    coord uc = {sq(sin(x0*pi))*sin(2.*y0*pi), -sin(2.*x0*pi)*sq(sin(y0*pi))};
    foreach_dimension()
      u.x[] = cdt*uc.x;
  }
  tTime += dt;
}

#if ADAPT
event adapt (i++) {
  adapt_wavelet ({mask_intf}, (double[]){0.02}, maxlevel = level, minlevel = level - 3);
}
#endif

event interface_out (i++, last) {
  if (2*(i + 1) % max(IT, 1) == 0 || i == 0) {
    int ii = 2*(i + 1)/max(IT, 1);
    char name[80];
    sprintf (name, "%s_ebit_%d_%d_%d.dat", OUTNAME, N, (int) EndT, ii);
    output_facets_ebit (name);
  }
}

    We can compute the shape error (E_{shape}) and area error (E_{area}).

    E_{shape}=\max_{i}| \mathrm{dist} (\boldsymbol{x}_i)| . \mathrm{dist}(\boldsymbol{x}_i)=\sqrt{(x_i - x_c)^2 + (y_i - y_c)^2} - R where the reference solution is a circle centered in (x_c,y_c) and with radius R.

    E_{area} = (A(T) - A(0)) / A(0).

    event calc_infty_norm (t = end) {
  double l_inf = 0.;
  coord dir = {0., 1.};

  foreach_face(reduction(max:l_inf)) {
    if (with_marker.x[] > 1.e-6) {
      double ss = (s.x[] - 0.5)*Delta, xx, yy;
      xx = x + ss*dir.x;
      yy = y + ss*dir.y;
      double dist = fabs(sqrt(sq(xx - xcenter) + sq(yy - ycenter)) - R);
      if (dist > l_inf ) l_inf = dist;
    }
  }

  // shape error and area error
  printf ("%d %e %e %e %e\n", N, area0, area, fabs(area0 - area)/area0, l_inf);

  // reference file
  output_facets_ebit ("", stderr);
}

    Results

    The shapes of the interface at t = T/2 and t = T are displayed below for both sets of simulations (T = 2, 8).

    reset
set size ratio -1
plot [0.:1.][0.:1.]'vortex_ebit_128_2_1.dat' w l lw 3 t "EBIT, t = T/2", \
  'vortex_ebit_128_2_2.dat' w l lw 3 t "EBIT, t = T", \
  '../vortex_ana_2_1.dat' w l dt 2 t "Ref. t = T/2", \
  '../vortex_ana_2_2.dat' w l dt 2 t "Ref. t = T"
    Shapes of the interface with period T = 2 (N = 128). (script) 
    reset
set size ratio -1
plot [0.:1.][0.:1.]'vortex_ebit_128_8_1.dat' w l lw 3 t "EBIT, t = T/2", \
  'vortex_ebit_128_8_2.dat' w l lw 3 t "EBIT, t = T", \
  '../vortex_ana_8_4.dat' w l dt 2 t "Ref. t = T/2", \
  '../vortex_ana_8_8.dat' w l dt 2 t "Ref. t = T"
    Shapes of the interface with period T = 8 (N = 128). (script)

    See also

    References

    [rider1998]

    William J. Rider and Douglas B. Kothe. Reconstructing volume tracking. Journal of Computational Physics, 141:112–152, 1998.