Eliminating Negative Volume Elements
Hello all -
Despite that I have used minimal smoothing in ICEM (I just smoothed for orthogonality, being that it is a hexahedral grid), my grid appears to have negative volume elements, as shown below.
Apparently, as shown on the left in blue writing and in the histogram, there are almost 63 negative volume elements (8 + 55), with the smallest being -18.7586 mm^3. I need to get rid of these, but am not sure how.
The ICEM Help file mentions that it does not consider "degenerate hex elements, and will assign a negative value for them."
Could this be the problem? If so, what is a degenerate hex element?
A hexa cell should have six surfaces, ie 8 nodes. A degenerate will have less, for instance, 7 nodes, or even 6 nodes (giving you a prism)
Thanks, Mark. I couldn't find an answer to that in the Help file or online. Upon closer inspection, I can see that there are some degenerate elements (e.g. the one shown below), as well as some negative volume elements.
I'm thinking the problem arises from the interaction of my expansive boundary layer on the blade surface and the larger structured H-grid preceding the blade, shown below (the second picture is zoomed-in to the stagnation point).
How could I make this interaction feasible? Maybe I should move the blocking vertexes slightly ahead of the stagnation point? Any additional input would be much appreciated.
Off the top of my head, I would suggest making a split in the blocks near to the surface of your blade that runs through the inlet and outlet sections as well. This would allow you to manipulate the shape of the block around the leading edge of the blade.
I have attached an image which shows what to do. Hope it helps!
Excellent suggestion by Mark,
Zooming in on the problem you can see that the mesh is tripping.
You could do a block split as mark suggested or simply split the edge like this...
Of course it doesn't help that your periodic boundaries meet the blade below the stagnation point...
you could also try smoothing. This is with the orthogonality smoother, the Multiblock smoother would be much better, but I am out of time for today...
Probably your best bet would be a different topology that included ogrids around the airfoil blade (Cgrids since you have only half a blade on each side)
I will try to get back to you with more later...
Hi Mark -
Thanks for the reply.
I assume that to perform that split in your picture, I should use the Split Block > Split Method > Curve Parameter method to allow the block to "follow" my geometry. Below is my blocking without the split.
The picture below shows the blocking with the suggested split (assuming I did it correctly).
Finally, here is the "finished" product blocking scheme.
With this scheme, my pre-mesh looks like this:
Hi Simon -
Using your method, I split the edge with Control Points located near the vertex. I obtained the following:
One thing I'm having problems with is using the Link Edge tool to make both block edges in the z-direction equal. In the above picture, I did both edges manually, and the overlay is pretty good, but I'd prefer it to be exact. The Link Edge tool, however, does not seem to overlay them properly.
The mesh looks like this:
I think this is the right idea. Obviously, I still have to refine the sizing and smooth with orthogonality.
To follow-up, before I've done any smoothing...
With the Split Edge method, I've obtained negative volume elements along the airfoil suction surface. In the picture below, you can see the mesh "colliding" with the airfoil surface, forming negative elements.
OGrids would avoid all this hassle.
Very strange... I will try to take another look some time this weekend...
I can probably block it out with Ogrids in a few minutes and save you some hassle.
Hey Josh... I spend a few minutes with it and created a 2D blocking (less work). You can always extrude this for CFX, or simply write it out as a 2D Fluent file and then read it into CFX (CFX will turn it into a 2.5D model).
I thought I recorded a replay script for you... but then I realized I had forgotten to hit record... Oh well.
Anyway... here are some pics.
First, here is the OGrid Setup (Blocks and Faces to select) for 3D...
Then here it is for 2D... (The Yellow blocks are VORFN blocks activated when I selected the edges)
Then I just used edge splits to get roughly what I wanted, and used edge params to get the distribution that I wanted... You can see that in the blocking file (sent privately).
Then I smoothed it with orthogonality and Laplace (and fudged a few nodes with "move nodes")
here are the final shots... If these won't run we can relax the periodic sides (so they can move but still be periodic) and get a better mesh.
Hi Simon -
You've been an incredible help. Thank you.
One thing, though. You mentioned you sent me (privately) the blocking file, but I never received it. Did you use my Gmail account?
Hi all -
I'm thinking of improving the quality of the post-blade mesh, as a quality check showed problems with the aspect ratio, skew, and overall quality in the angled region just after the blades. You can see the skewing quality in Simon's two bottom pictures in his last post. As shown, the quality could be improved.
I'm thinking there are two ways to go about this:
1) Adjust the blocking (edge distributions, vertex locations) in that region. This seems like the better option, though I'd have to re-mesh and, thus, eliminate all of the fine smoothing and node adjustments Simon performed.
2) Smoothen for quality. I know this isn't a great idea, especially for hexa meshes, and could cause overlap and degenerate/negative elements.
What do you think?
Also, just to follow-up: Simon's mesh ran perfectly in CFX.
Go for it.
Sure, go for it, but keep in mind that you need a periodic model... This makes it difficult because you will have to deal with sloping channel behind the airfoil.
Sorry, I haven't had time to do much more than type a quick message. My machine is cranking right now and I just can't run any more tasks...
For smoothing, I just used orthogonal for a bunch of iterations and a few iterations of laplace, then back to orthogonal for 10 more. Then I used move nodes to straighten up some of the nodes ahead of the lower airfoil... Perhaps it would have been a bit better if I had relaxed the orthogonality requirement on the periodic walls...
You could try a "shifted periodic" blocking...
Thanks, Simon. I'll see how I'm doing for time.
I hope you are well. I cam across your post just the other day when i was looking for blocking strategies for a turbine blade i am modelling. I usually, use the blade-in-the-middle approach with an o-grid, but for specific reasons i need to model it with the blade surfaces at the outer edge, like you did here.
however, having read your posts, i was wondering how / whether, you managed to get a fully periodic (cyclic) mesh out using this approach i.e. did you manage to run your simulations ok using this mesh.
i know it is only 2D, but i assume you needed to set periodic vertices and the periodic rotation angle in the global mesh setup? Would you be willing / able to give me an idea of how you did it?
many thanks and best regards
|All times are GMT -4. The time now is 22:02.|