Author |
Message |
Bernd Vosseler (bernd)
New member Username: bernd
Post Number: 1 Registered: 02-2006
| Posted on Wednesday, February 01, 2006 - 11:57 am: | |
I am using Matlab to post-process data gained by a thermal simulation. To Export the data, I first tried using the EXPORT function and imported the file to Matlab with the fpreadsc routine posted here. The result was very noisy, compared to the plot displayed by FlexPDE. I then tried exporting the data by generating evaluation plots along lines of constant y, but some of these plots were empty domains although the path was correctly addressed. How can I avoid noisy 3d-output and/or empty plots? 60.6 K | Surf_T_export.pg5 "Surface Temperature Calculation" | | |
Jerry Brown (jerrybrown11743)
Member Username: jerrybrown11743
Post Number: 18 Registered: 03-2004
| Posted on Wednesday, February 01, 2006 - 01:09 pm: | |
Judicious selection of contour levels might help. I'm not familiar with Matlab's capabilities. But, I can recommend Tecplot. It gives you complete control of the individual contour levels. |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 528 Registered: 06-2003
| Posted on Wednesday, February 01, 2006 - 07:19 pm: | |
When FlexPDE creates a plot, it subdivides the computation mesh (using the current interpolation polynomials) until the desired resolution is achieved (default for contour plots is 51x51; control the density with SELECT CONTOURGRID=# or the POINTS=# qualifier on the plot command). EXPORT overlays the plot area with a rectangular grid and exports the interpolated data on the grid. This causes a loss of precision as compared to the FlexPDE native plot. The resolution of the EXPORT is controlled by the same values as the FlexPDE plot resolution (ie, the POINTS qualifier also controls the density of the exported table). You should be able to improve the quality of the offline plots by increasing the EXPORT density. Try POINTS=100 or 150. TECPLOT and VTKLIN export formats do not refine the computation mesh, but convert it to a linear triangular or tetrahedral mesh (eg, 8 linear tetrahedra per quadratic computation tetrahedron). This results in a loss of smoothness, which you can compensate by using a smaller ERRLIM and/or explicit MESH_SPACING commands. |
Bernd Vosseler (bernd)
New member Username: bernd
Post Number: 2 Registered: 02-2006
| Posted on Thursday, February 02, 2006 - 04:14 am: | |
Hello Robert, using points=(321,241) improved the plot Quality (surf4), but the result is not as smooth as the one I yielded combining many elevation-on-path outputs. Plus, I got a very smooth output from an internal surface (surface 2) of the problem using the same output method as for the plot displayed at top. So I think it is neither a resolution too small, errlims too large, nor a question of Matlab's capabilities. Something ill seems to happen when exporting the top surface. I managed avoiding empty plots in the elevation-on-path plots by simplyfying the problem to only one layer. Do you know, where that problem comes from?
|
Bernd Vosseler (bernd)
Junior Member Username: bernd
Post Number: 3 Registered: 02-2006
| Posted on Thursday, February 02, 2006 - 06:38 am: | |
Sorry, it seems I could not properly format my message and the image annotations. From left to right, the first image is the one I generated from multiple ELEVATION plots, the second (noisy) one is the top surface as exported with SURFACE, the third one is an internal one exported with SURFACE |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 535 Registered: 06-2003
| Posted on Thursday, February 02, 2006 - 07:41 pm: | |
I have run some tests, and it appears there is indeed an error in EXPORTED data on 3D cut planes. I will investigate and correct this as soon as possible. |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 536 Registered: 06-2003
| Posted on Friday, February 03, 2006 - 03:09 pm: | |
I have discovered the error in file export that caused your difficulty. (It is a software-rot issue involving definition of the cutplane normal and the ordering of cell nodes. It should affect only cutplanes at constant Y coordinate). The correction will appear in the next maintenance release (5.0.10), but you can download a development copy for Windows from www.pdesolutions.com/download/xfpde5010x2.exe. This is a self-extracting archive containing only flexpde5.exe; it is not a full install. Extract the file in place of your existing flexpde5.exe, or change the name of the old file first.
|