1-D Riemann tests

The 1-D L&W tests consist of single Riemann problems with known solutions, and an interacting two shock test with no analytical solution, but which may be used to compare low- and high-resolution results.

NOTE: The final time for the Blast problem was omitted by L&W, but has been determined from previous published results and is presented here.

The code and configuration files plus output for Fyris to run the 1D Riemann problems will be available soon.

The 6 simple tests

These are specified as follows:

  • adiabatic index: Gamma = 1.4
  • grid domain: 0.0 < x < 1.0
  • all cells with central coordinates < x0 are set to the Left State, all cells with central coordinates > x0 are set to the Right State (x0 lies on an exact cell boundary)
  • tests are run from time = 0 until time = T
  • left and right boundaries are free or natural boundaries that replicate the adjacent grid values.
  Left State
Test Density Pressure Velocity
1 1.0000 1.0000 0.7500
2 1.0000 0.4000 -2.0000
3a 1.0000 1000.0000 -19.59745
4 5.9992 460.894 19.5975
5 1.4000 1.0000 0.0000
6 1.4000 1.0000 0.1000
  Right State     Grid
Test Density Pressure Velocity x0 T Cells
1 0.1250 0.1000 0.0000 0.3000 0.2000 100
2 1.0000 0.4000 2.0000 0.5000 0.1500 100
3a 1.0000 0.0100 -19.59745 0.8000 0.0120 200
4 5.9924 46.0950 -6.1963 0.4000 0.0350 200
5 1.0000 1.0000 0.0000 0.5000 2.0000 100
6 1.0000 1.0000 0.1000 0.5000 2.0000 100

The 1D Noh Test

  • adiabatic index: Gamma = 5/3
  • grid domain: 0.0 < x < 1.0
  • run until time = 1.000
  • free boundaries
  Left State Right State     Grid
Test Density Pressure Velocity Density Pressure Velocity x0 T Cells
1 1.0000 1.0e-6 1.0000 1.0000 1.0e-6 -1.0000 0.5000 1.000 100

The 1D Peak Test

  • adiabatic index: Gamma = 1.4
  • grid domain: 0.1 < x < 0.6
  • run until time = 0.0039
  • free boundaries
  Left State Right State     Grid
Test Density Pressure Velocity Density Pressure Velocity x0 T Cells
1 0.1261192 782.92899 8.9047029 6.591493 3.1544874 2.2654207 0.500 0.0039 800

The 1D Woodward-Collela blast wave test

  • adiabatic index: Gamma = 1.4
  • all cells with central coordinates < x0_left are set to the Left State, all cells with central coordinates > x0_right are set to the Right State, other cells are set to the middle state.
  • left and right boundaries are reflecting boundaries
  • tests are run from time = 0 until time = 0.038
  • grid domain: x = 0 -> 1
  Left State Middle State Right State
Test Density Pressure Velocity Density Pressure Velocity Density Pressure Velocity
1 1.000 1000.0 0.000 1.000 0.0100 0.000 1.000 100.00 0.000
        Grid
Test x0_left x0_Right T Cells
1 0.100 0.900 0.038 400/2000
 

Results

  • The models are run and compared with the exact solution, or the 2000 cell model for the blast wave test.

  • The exact solutions are integrated over each cell to get cell average values at the same resolution as the test models.

  • For the exact solutions the cells were subsampled by a factor of 5 and averaged, consistent with the factor between the 400 and 2000 cell blast tests.

  • L1 norms between the models and the integrated exact solutions are computed as percentages.

  • All L1 norms were performed on the density, except for Test 2, where it was on Specific Energy = P/(d*(gamma-1.0))

  • The exact solutions were computed with the exact Riemann solver code riemann.c (see also the Riemann page).

  • The exact solutions are included the riemann.c download as a set of text files.

  • All Fyris Alpha 1-D tests were run with a Courant number of 0.8, with an initial step with a Courant number of 0.64 (0.8x0.8).

The Fyris Alpha L1 norms for the tests 1-6, Noh, Peak, and Blast are shown below in a table that includes the results from Liska and Wendroff. Each is the L1 norm as a percentage for density, except Test 2 which is internal energy, and Peak which is the velocity L1 norm. Blast is given as the L1 density norm percentage w.r.t. the 2000 cell solution Blast density, rebinned by a factor of 5. To compute the other L1 norms, the exact solution was evaluated in 5 locations across each cell and averaged, allowing for solutions to have boundaries part way through cells, and remaining consistent with the rebinning used for the Blast L1 norm.

The results are similar to PPM and VH1, which use related algorithms, although Fyris does significantly better with the strong shock Noh problem. This is reflected in the 2D tests where Fyris also performs the difficult 2D version of the Noh problem very well, using standard settings, and with sufficient stability to derive an L1 norm for density and pressure for the 2D case. The smallest error code for each test is shown in red.

 
Code 1 2 3a 4 5 6 Noh Peak Blast
Fyris 1.0 9.8 3.6 1.3 0.0 0.3 0.95 0.8 5.3
CFLFh 1.5 10.2 10.3 2.7 0.7 0.8 1.9 1.9  
JT 1.3 6.4 8.1 2.3 0.6 0.6 1.7 1.1  
LL 1.3 31.3 5.2 2.4 0.5 0.7 1.5 0.8  
CLAW 0.8 fail 3.1 1.7 0 0.4 1.3 fail  
WAFT 0.7 21.9 2.6 1.4 0 0.3 2.8 1  
WENO 1.3 23.7 9.2 2.2 0 0.4 2.0 2.4  
PPM 0.5 6.3 9.4 1.1 0 0.1 4.6 1.3  
VH1 0.9 9.6 3.7 1.3 0 0.3 1.5 0.8  

Plots comparing the exact and computed models follow. Finite models are shown as yellow dots, exact solutions as red line and red square dots.

Test 1 density

TEST 1 DENSITY

Test 2 energy

TEST 2 ENERGY

Test 2 density

TEST 2 DENSITY

Test 3a density

TEST 3A DENSITY

Test 4 density

TEST 4 DENSITY

Test 6 density

TEST 6 DENSITY

Test Noh density

TEST NOH DENSITY

Test peak density

TEST PEAK DENSITY

Test peak velocity

TEST PEAK VELOCITY

Test peak density zoom

TEST PEAK DENSITY ZOOM

Test peak velocity zoom

TEST PEAK VELOCITY ZOOM

Test blast density

TEST BLAST DENSITY