open TELEMAC-MASCARET The mathematically superior suite of solvers

Thanks again Chris.

Still stuck - so to answer your questions;

1. Yes, there is a temp directory that contains:
a. an almost empty config file (only two numbers)
b. an empty sortie file
c. a copy of my .cas file (t2dcas)
d. a copy of my .clm file (t2dcli)
e. a t2ddico file (looks complete)
f. a t2dgeo file (that I can open just fine in BlueKenue)
g. a t2dres file (unsurprisingly empty).

2. The code stops dead after copying the files. No helpful error message. I think it must relate to my particular model, because I can run the test cases fine, and have managed to run a few simple made up models created in BlueKenue as well.

3. I'm reticent to send the files over the forum for confidentiality reasons - but could send them directly.

Also, from my investigations today I am now fairly certain that I must have a problem with my mesh (and a related problem with the .clm file). The mesh is huge and complex (not a good place to start as a beginner user). I have around 300,000 nodes and 550,000 elements that describe the rather complicated coastline of British Columbia. I have checked the mesh for as many things as I know to check for (more than 8 neighbours, less than three elements between boundaries, nicely shaped triangles, etc....I've been building 2D meshes for 15 years - just not for TELEMAC). I've been using TRIGRID to work on most of these issues. Though, according the the mesh analyser in BlueKenue I have 1 edge node with 9 neighbours that doesn't appear in TRIGRID (I've also written a macro to search through the .t3s file to no avail). Is there a way to track down this errant node/element? Also, at one point I got TELEMAC to go one step further (it crashed after sorting out memory for the mesh) and it provided an error message that related to a specific element (not the node) - is there an easy way to find an element in BlueKenue, TRIGRID or anything else? I eventually tracked down the element by looking for the proverbial needle in the haystack, and there was in fact an error in the .clm file at this point (it looked to be a bug in the CONLIM creator program in BlueKenue - I'll describe this separately in the Pre/Post Processor forum once I'm sure of what might have happened).

So, to summarise - I think I'm dealing with errors in my mesh, but I'm having a hard time figuring out what they are.

Thanks again for all your help to date (and of course for any future help),

Tamsin

Thank you for the detailed answer. It gave me a few more ideas about what might be going wrong.

It might be that your IPOBO array has become corrupted somehow, perhaps if you've been editing boundary nodes by hand (even in BlueKenue). Here's how to check/fix this:
1) Load your geometry file in BlueKenue
2) View the mesh in a 2D View
3) Drag the IPOBO variable into the same 2D view.
4) Look at the points that appear. Make them bigger using the properties of the IPOBO object if it helps. If any of them are not on the model boundary, then that's your problem -- IPOBO points should only be on the boundary.
5) If your IPOBO looks wrong, create a new selafin object and drag your mesh into it. This should calculate a correct IPOBO array automatically.
6) Save your new selafin object, and try TELEMAC again (and cross your fingers).

Another possibility is that you could have an isthmus in your mesh (i.e. a single node connecting two otherwise unconnected components of the mesh). I suspect that you've already checked for this, but since it doesn't show up in the BlueKenue analysis I thought I'd mention it. I'm afraid I don't have a fast way to check for this. You can brute force it by starting with a random element and growing a component through finding triangles it shares an edge with, but it's very slow going. There's probably a more cunning method out there.

I'm puzzled by your edge node with 9 neighbours. I guess it could be a symptom of a broken IPOBO array (see above), but I'm not convinced of that.

Finding an element is easy in BlueKenue. Just right-click the mesh, and go to Select -> Element ID, then enter the element you're looking for. This will select the element. Spotting it in a complicated mesh is a matter of keen eyesight. Good luck!

If you want me to take a quick look at your geometry file, then you can send it to me by email at c.cawthorn <<at>> hrwallingford.com.

Thanks Chris.

I think you may have hit on the problem. I've spent the day comparing the various input files - .ipobo to .clm to .ngh to the resampled boundary lines and have identified about 20 spots on the boundary where the .ipobo and .clm do not match the boundary (.ngh and resampled boundary) - so essentially there are points on the boundary that have a node code of 0. Unfortunately for me - these didn't show up when I completed all the tests in TRIGRID.

Ultimately, I think the problem relates to BlueKenue (you may want to read the post from my colleague 'ingenieure' in the pre/post-processor forum). We were having problems, I think, at sites where the distance over dry land was less than the density of the mesh at that point. I thought I'd caught all of these errors and adjusted the density - however, given that I've got thousands and thousands of kilometres of coastline apparently I missed a few. At this point I'm going to work my way through all of the points I've identified and fix the mesh as required. Hopefully, this will at least get the model going (though I'm not naive enough to think that it will produce results just yet....).

Thanks also for the tip about finding elements in BlueKenue, I had been using the select item feature under the edit menu, which only allows you to select nodes.

Thanks again,

Tamsin