Changes between Version 66 and Version 67 of u/erica/2DShockedClumpsSNR

Timestamp:

01/31/18 09:07:44 (7 years ago)

Author:

Erica Kaminski

Comment:

—

u/erica/2DShockedClumpsSNR

@@ -69 +69 @@
 This run was just to see if the choice of solver can significantly change the simulation. To check this, here is a line plot of the density vs. position for the PPM method (the previous simulation used PPL): 
 
-[[Image(density_front.png, 50%)]]
+[[Image(density_front.png, 40%)]]
 
 
 and a comparison of the final frame:
 
-[[Image(Last_Frame_Comparison.png, 75%)]]
+[[Image(Last_Frame_Comparison.png, 65%)]]
 
 === Run 3 - Pulsed case=== 
@@ -115 +115 @@
 
 
-[[Image(amr_mesh.png, 75%)]]
+[[Image(amr_mesh.png, 65%)]]
 
 In testing AMR on this setup, I was finding that the meshing algorithm produces slightly different AMR patch distributions when different numbers of processors are used (for reference, the run presented here used 16 processors). It seems that numerical errors are produced within the clump when the mesh is asymmetric, resulting in asymmetries in the fluid instabilities (compare left image with AMR to right without AMR):
 
-[[Image(AMR_Fixed_Comparison.png, 75%)]]
+[[Image(AMR_Fixed_Comparison.png, 65%)]]
 
 This is particularly noticeable along the central axis of the clump, normal to the shock. More on this in the conclusions section below.
@@ -126 +126 @@
 === Run 7 - Prelim 3D run ===
 
-[[Image(3D_contour.png, 75%)]]
+[[Image(3D_contour.png, 65%)]]
 
-[[Image(3D_Mesh.png, 75%)]]
+[[Image(3D_Mesh.png, 65%)]]
 
 Estimate ~ 5 hours per frame with cooling turned on at same resolution above (on 24 processors).
 
 === Run 8 - Conclusions ===
+
+
+The above results show that the structure of post-shock instabilities that form along the surface of the clump can be sensitive to the choice of solver as well as the alignment of the mesh with respect to the clump center. This suggests that we should break the symmetry in the initial conditions to avoid (unphysical) behavior based on numerical error. One way to do this would be to introduce a sinusoidal perturbation to the shock front.
+
+In addition, we would like to experiment with the internal density distribution of the clump, e.g. using a Gaussian density profile instead of a uniform distribution, as well as remove the pressure equilibrium constraint on the clump/inter-clump medium, and add a cooling prescription to the simulations. Lastly, there is no reason to not initiate the supersonic flow within the computational domain as well as along the boundary. Initializing the shock just a few cells away from the clump will minimize computation time as well as numerical errors as the shock propagates along the grid.
+
 
 === Unresolved thoughts: ===