error during mesh generation

A discussion forum for users of the new HYDRUS 2D/3D. HYDRUS is a software package for simulating water, heat and solute movement in two- and three-dimensional variably saturated media. Happy Posting!
Post Reply
Luisa
Posts: 35
Joined: Tue Mar 06, 2007 8:22 pm
Location: USA

error during mesh generation

Post by Luisa » Thu Oct 23, 2008 4:27 pm

Hi,

when I was trying to generate a 3D mesh I got the error message:
"Internal error no. 62 during generation of 2D FE-mesh! Mesh on Surface no.1 is inconsistent."

How can I check where or what the error is and correct it? I have to admit that I am trying to build a rather large, irregularly shaped domain with 2 sublayers (small watershed) by importing a DEM.

Thanks, Luisa

Mirek
Posts: 417
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Thu Oct 23, 2008 6:28 pm

Hi Luisa,

This error can be caused for example by two boundary points with almost identical coordinates, i.e. they look like a single point but there actually are two different points. Their distance can be extremely small, which can cause problems during FE-mesh generation.

I suggest that you should:

1/ Check duplicity of your boundary points (if the number of points is not too large): Select each point by the rectangle and pres Alt + Enter to display the “Edit Points” dialog. Check that there is the only one number in the field “Point No.”, otherwise you have overlapping points.

2/ Download and install the latest version of HYDRUS (1.07) which contains several improvements and fixes of the FE-mesh generator http://www.pc-progress.cz/Fr_Hydrus3D_Downloads.htm

3/ If your problems continue, please, send me your project, I’ll look at it.

Regards
Mirek

Luisa
Posts: 35
Joined: Tue Mar 06, 2007 8:22 pm
Location: USA

Post by Luisa » Thu Oct 23, 2008 6:58 pm

Hi Mirek,

I am using version 1.07 and will try your first option although I do have a lot of boundary points. ;-) I will get back to you if I can't figure out the problem.

Thanks a lot! Luisa

Liesbeth
Posts: 1
Joined: Fri Feb 05, 2010 4:57 pm
Location: Belgium

Post by Liesbeth » Mon Feb 08, 2010 10:06 am

Hi,

I have some problems with the Fe-Mesh generation for my 2D-project (with a general domain definition). I always got the error message of internal error nr. 70 (the mesh on surface no. 1 fails to satisfy the model precision criterion…)

The project concerns a 2D-transect of +/- 1000 m length and a depth of 120 m. The upper surface layer follows the topography and contains a node every 5 m, exept at the side of the river, where we added nodes every meter over a distance of 100 m. This line (upper surface layer) is copied 10 times and placed at 0.1 m distance of each other in order to specify the root distribution precisely over a distance of 1m in the upper layer. We also specified some additional nodes (1 m distance) at the side boundary lines in the upper layer (over a distance of 5 m). For the rest, there are only nodes at the corners of the lower surface layer. At the river side, mesh refinement (S=0.1) was inserted at the boundary nodes and some extra nodes which are lying just near the boundary.

I tried already to change the mesh-parameters like the stretching factor, the smoothing factor,… but with no result. Lowering the epsilon value (which is not recommended) gave already better results, but I still get the error message. Also a simpler version of the project (without copies of the upper layer) could not help to solve the problem.

The strange thing is that similar projects were run a year ago by someone else without any problems. But if we remove the mesh in these ‘old’ projects and try to generate again a mesh, we got the same error message.

Any idea where the problem is situated?

Thanks in advance,
Liesbeth

Mirek
Posts: 417
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Mon Feb 08, 2010 10:15 am

Hi Liesbeth,

Please send me the project (mireks@pc-progress.cz), I'll look at it. Use WINZIP to pack it. What version of HYDRUS do you have, 1.10?

Regards Mirek

Mirek
Posts: 417
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Tue Feb 09, 2010 9:44 am

Hello Leisbeth,

The problem is that your domain is about 1km long and at the same time it contains some geometrical details with resolution about 1cm. This means that the ratio Dmin/Dmax = 1/100000, where Dmax = diameter of the whole domain and Dmin = diameter of the smallest finite element (= diameter of the inscribed circle). FE-mesh generation in HYDRUS is done in single precision, i.e. the maximum resolution ratio is 1/1000000. As you can see, your geometrical model is very close to this limit and this is also the reason why FE-mesh generation fails in your project. The attached figure explains the situation – the circle inscribed in the yellow triangle has diameter D = 1cm. Although this triangle is just temporary (i.e. it will be replaced by some less-flat triangles in the final mesh), it can cause problems in the FE-mesh generation process.

Solutions:
1/ Although it is not very easy to generate a FE-mesh for your domain, it is possible. I will send you the finished mesh and some comments and hints today.

2/ I have an idea how to extend the precision by 1 order, which should be sufficient for your project. If it works, we could prepare an update (version 1.11) within 1-2 weeks.

3/ The final solution of this problem will be in using double-precision for the FE-mesh generation calculations. This is planned in version 2.xx.

Regards
Mirek

Image

MOD_LAB
Posts: 14
Joined: Wed Feb 10, 2010 6:56 pm
Location: Tunisia

Post by MOD_LAB » Thu Feb 18, 2010 10:23 am

Hi Mirek,

I was trying to continue with the old file (without translating the domain next to the origin). I used the "Repair geometry function" and I continue the steps of the tutorial 2.07. I have done successfully the construction of the solid. However, I have an error during the mesh generation ("Internal error no. 62 during generation of 2D FE-mesh! Mesh on Surface no.1 is inconsistent"). I read that it is due to extremely small distance between points. I want to know if this problem is related to the position of the domain (being far from the [0 0 0]). I am working with 1.10 version of Hydrus.
What do you think Mirek??

Thanks in advance

Regards

Mirek
Posts: 417
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Thu Feb 18, 2010 10:56 am

Hi Maroua,

Just for the others: your question is actually regarding another topic in this forum – see
http://www.pc-progress.cz/_forum/topic. ... IC_ID=1425

The problem with the FE-mesh: Your domain is defined in global coordinates (e.g. x=587012.594 m , y=397585.9776 m), which is not good. This can cause many problems, including the problem with the FE-mesh accuracy. You need to move your domain to the origin [0,0,0], which can be done by several ways. For example:
1/ Import your X,Y,Z coordinates to MS Excel
2/ Create two new columns for the transformed X,Y-coordinates
3/ Calculate X0,Y0 coordinates using a simple formula X0=X-587012, Y0=Y-397585
4/ Export columns with transformed coordinates to a text file
5/ Prepare the input file for HYDRUS (see HYDRUS help - Graphical User Interface -> Menu -> File -> Import and Export ->Import Geometry from a Text File, KEY_WORD: THICKNESS_ARR3Z_NLAYERS ).
6/ Import the coordinates in HYDRUS

If your domain is located at the origin, the FE-mesh should be generated without problems.

By the way: The precision of the FE-mesh generator has been improved and now we are able to generate the FE-mesh for Liesbeth’s projects. This improved generator will be available in version 1.11, which will be released officially next week.

Regards Mirek

erwin
Posts: 9
Joined: Mon Aug 12, 2019 7:43 am
Location: Germany

Re: error during mesh generation

Post by erwin » Mon Mar 02, 2020 4:52 pm

Dear Mirek,

This is an old thread. But I want to ask what is the maximum resolution now for the HYDRUS 3.01? I am doing a 2D model of a coastal peatland with a transport domain length of 2000 m and a height of about 12 m. I want to have a resolution of 0.1 m (10 cm) as this is the most interesting for peatlands. Based on these measurements, my resolution would be 1/20,000,000.

I have been stuck on finalizing my general model since I can only have 0.11 m resolution in the surface if my transport domain has a 150 m length and 12 m height (target size is 0.6 m, refinement of 0.2 m). I want to have a transport domain that would extend to 2000 m.

Thank you in advanced for your answer!

Best,
Erwin

Jirka
Posts: 4861
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: error during mesh generation

Post by Jirka » Mon Mar 02, 2020 5:29 pm

Since the pressure head isolines in your problem are likely horizontal, you should use an anisotropic FE mesh. For such conditions you can easily use an anisotropy ratio of the mesh of 1:10, 1:100, or even more. You need to test it. Also, note that you need to have finer discretization only in the unsaturated zone and much coarser mesh in the saturated zone. J.

erwin
Posts: 9
Joined: Mon Aug 12, 2019 7:43 am
Location: Germany

Re: error during mesh generation

Post by erwin » Wed Mar 04, 2020 7:21 am

Dear Jirka,

Thank you so much for the suggestion! I was able to make a 2000 m domain with the height of a node at ~0.03 m with an anisotropy ratio of 100 using the stretching factor function.

I tried doing a 1000 m domain with stretching of X=100 and X=10. The streamlines are different and I seem to be getting a difference of one order of magnitude in the sea-side boundary fluxes. I don't think I should be getting that larger difference if I changed the stretching factor.

Thanks, Erwin.

Jirka
Posts: 4861
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: error during mesh generation

Post by Jirka » Wed Mar 04, 2020 6:30 pm

As I wrote above you need to test different grids, with such a large grid anisotropy. You should look at the mass balance errors and accept results that gives you good mass balance errors. J.

erwin
Posts: 9
Joined: Mon Aug 12, 2019 7:43 am
Location: Germany

Re: error during mesh generation

Post by erwin » Mon Mar 09, 2020 8:49 pm

Dear Jirka,

Thank you for the recommendations. Different simulations using stretching factors of X=10 and X=100 have mass balance errors of less than 1%. However, the stretching factor X=10 gave a more realistic simulation than the X=100. There was a big difference in the generated streamlines and boundary flux values. The simulation with a higher number of nodes in the X=10 looked more realistic than what was generated from X=100.

Best,
Erwin

Jirka
Posts: 4861
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: error during mesh generation

Post by Jirka » Mon Mar 09, 2020 9:36 pm

I would not care that much about streamlines (and velocities) since that is simply postprocessing of results by GUI. I would care more about the calculated pressure heads and boundary fluxes, which are a direct result of calculations. J.

Post Reply