Inventor Nastran performance issue

[sprotsenko]

Hi, I am stress analyzing strainer basket yet there are performance issues.

This is screenshot of basket. It is literally cylinder with holes on it. I applied internal pressure to see the stress.

However I was not able to generate any mesh since it takes forever to do so.

I started reducing model shortening it, it does not help.

I ended up with shorter model and only 5 deg sector of the holes. I even tried to make 15 degrees, yet it became frozen when generating the mesh. It does not take more time to calculate the mesh, it is just frozen. I was waiting for 1 hour and gave up. 5deg sector was calculated in several minutes.

I attached this model file.

Definitely there is a breaking point with the number of holes when Nastran is stuck with mesh calcs.

If there is any solution for this it would be great.

[John_Holtz]

3016-BASKET RING meshes very easily.

For the model with the holes, I think you should fine a theoretical paper that describes how to model a "screen" or "sheet with a million holes". Trying to mesh something with that many holes is unreasonable because the mesh size (number of nodes and elements) will be huge. Hopefully another reader will know of such a paper since similar questions have come up in the past. For example, the IPA (Industrial Perforators Association) has "Designers, Specifiers and Buyers Handbook For Perforated Metals" that may have something useful.

If you need the real detail at some location, then include 20 holes at that area and use the theoretical paper to get the proper stiffness for the rest of the model of million-20 holes.

John

[sprotsenko]

Sorry. It was wrong step file. Original file is 100Mb. Forum has 70 Mb limit.

I will try to ZIP it.

[sprotsenko]

This is reduced basket file. If you make 15 deg sector, it will stuck. At least my Nastran.

[sprotsenko]

I understand that more hole make create more entities yet it is not more elements. if you cut the volume so number of element will be lower if we use the same size of elements.

As I mentioned in email the problem is when I create 3 times more holes (15 deg sector), it does not want to calculate. 

5 deg sector takes several minutes to calculate. QTY elements is not the issue here. Something wrong with software when it has limit and does not want to calculate after some number of holes.

[John_Holtz]

Hi @sprotsenko 

Sorry, I jumped over the meshing question and was answering a different question about running the analysis.

Yes, meshing will be slower when you have more features lines. The model shown is very thin; can you use shell elements instead of solid? That may reduce the time to create the mesh since the number of feature lines would be cut in half. Also, solids require a volume mesh, so there is additional time to create that which would not be necessary for a shell mesh.

Here are some times I measured creating shell meshes using the outside face and using Inventor Nastran 2025. (I used version 2025 partly because it has some mesh diagnostics that I wanted to try if the model failed to mesh. It meshed okay, so I didn't get to try the mesh troubleshooter.)

Number of holes	3 inch mesh size	1 inch mesh size	0.1 inch mesh size
0	1 seconds	1 seconds	16 seconds
90	28 seconds	28 seconds	42 seconds
270	230 seconds	240 seconds	245 seconds

I then used version 2023.2 for the case of 90 holes, 1 inch mesh size. That took 37 seconds to mesh using version 2023.2 (compared to 28 seconds with version 2025). Let us know if your computer is much slower than the times above. If so, I am not sure what the cause would be.

For your small test model, the ring is 40 inch diameter and 12 inch high. If there were no holes, a 3 inch mesh size is not too bad (depending on what you are trying to do). That would give approximately 160 quad elements. With the holes, you are trying to mesh at a size of 0.1 inch! (And that is probably too large, depending on what you are trying to do.) How many elements will it take to mesh the test model with a mesh size that small?

My original point was that the analysis will take much longer when you include the million holes because the number of nodes and elements will be 1000 times more than you need compared to a procedure to represent the piece without using any holes. Sure, it takes time to find the procedure and do the calculation of how to make it equivalent.  (Is it just changing the modulus? thickness?) But you will probably run the analysis multiple times, so you need to include how much extra time is spent in each analysis, and viewing each result, with a large model with holes compared to a small model without holes and time spent to calculate the equivalent model.

I have attached the Perforated Metals Handbook in case you decide to investigate that approach.

John

[sprotsenko]

Thanks for reply.

The main problem was not in processing time with model that I attached. 

If you modify it a bit increasing number of holes and run calcs, it will be frozen.

This is existing model that is ok.

If I try more holes, computer will be frozen when calculating stresses.

Ideally I would like to have 360 deg yet had to use 5 to run calcs.

The problem is not in number of elements. Model above with 5 deg sector works with any element size. Ones number of holes increased, it stops calculating and comp is frozen.

[sprotsenko]

This is my computer parameters

I did not try shell FEA element. I have to investigate it since I have never used shell Element in Inventor Nastran.

[John_Holtz]

Hi @sprotsenko 

Hi Sergii,

To clarify for other readers, the computer becomes frozen when generating the mesh, correct? When you write "and run calcs", " frozen when calculating stresses", and "it stops calculating and comp is frozen", that implies that you created the mesh successfully and were running the analysis. Generating the mesh and running the stress analysis are two completely different things. (Perhaps you found a way to delay the mesh generation until you run the analysis. Most users generate the mesh before trying to solve the analysis. I generally create the mesh, apply loads and constraints, run the analysis.)

I agree that Inventor Nastran cannot mesh the model when the number of holes gets larger and larger. It seems like your option is to find a different approach that does not use holes, or do not perform the analysis.

Did you time how long it takes your computer to mesh the model and compare your times to my times? Just curious if there is a large difference or not. 

John

[sprotsenko]

These is my time in seconds.

When I RUN calcs it is frozen. Mesh is preliminary generated.

It is probably takes so much time to get results so I gave up. I left computer over lunch time yet when I came it was frozen (may be calculating), so I killed the process in task manager. It was 1.5+ hours calculating time with 1" mesh and 270 holes.

[sprotsenko]

My assumption was that model with holes calculations should take less computing time then model without holes since number of elements is less in model with holes. Yet, it looks like this is not the case.

[John_Holtz]

Hi

I think this is the model that you provided on May 1. I added more holes to get 270 holes (= 15 holes in 18 columns).

• Meshing requires 333 seconds.
• The analysis elapsed time is 32 seconds. (25 seconds actual run time, other 7 seconds to generate the Nastran file and get started.)

Is your model saved on a network drive? (If use Windows File Explorer and go to the folder where the Inventor file is saved, what is the drive letter? Also from Inventor's "Model" tree, right-click on the model name at the top of the tree and choose "Open File Location". If the file is not saved to the C: or 😧 drive, it is probably on a network drive. Communicating over a network is slow, but it should not change the analysis from 33 seconds to hours.)

John

[sprotsenko]

Thanks John. Your results are promising. I will try again and investigate tomorrow. Thank you again.

[sprotsenko]

I checked the file, it is on C drive. My C drice is SSD so it should be fast.

The main problem that I see that Nastran has braking point when after certain amount of holes in shell it is frozen or calculation time is so long that I interrupt the process (typically after 10-15 min).

I tried several different ways to make it. It works for 22 degree sector. See below. I made shell element.

If I make 45 or 90 degree it is stuck with Mesh. 

I would like to have a least 45 deg to have basket representation. I will be running Linear buckling so holes matters.

Nastran should have status bar with expected calculation time. Status bar that it has does not show anything.

[sprotsenko]

For instance I have this progress bar when making 2" shell mesh. 

I just modified model and generating mesh. Yet no progress. It is frozen, I have to end the process in Task manager after 20 min since I do not know how much time it is going to take and I do not have time to wait for 4 hours and not to get result at the end.

[sprotsenko]

Update. I left computer over the lunch and it took 40+ minutes to make a mesh. I run Linear bukling and it took seconds to get a results. It looks like meshing algorithm is not ideal. It takes too mush time for relatively simple mesh. 

[John_Holtz]

Hi @sprotsenko 

I agree that there is something wrong with the meshing. My guess is the "round holes" through a cylindrical surface create splines on the surface (saddles), and it is the splines that are causing the problem with meshing.

Regardless of what causes the issue, meshing is not going to work. Continuing to try is pointless, so it is time to try something else.

1. Your part is thin, so you should not be creating a solid. A solid has twice as many splines which doesn't help. You should create a surface.
2. For your purposes, what is the difference between a "perfect" cylinder and a polygon with 144 sides? (That is 360 degrees divided by 2.5 degrees for two vertical columns of holes.) Is the difference 1%? 10%? 50%? (It is easy to test a model without holes.) It might take some time to mesh a polygon with 144 sides with perfectly circular holes, but I suspect that it will be much faster. (It is faster. See Figure 3.)
3. Buckling only depends on the stiffness. Do the holes really matter as long as the stiffness of a solid shell is the same as the shell with holes?
4. If you want to get really crazy, create your own mesh. Mesh the plate around 2 holes (or N holes, doesn't matter). Copy the mesh to the other locations, tangentially and axially. See Figures 2 and 3. It still requires a lot of work but may be possible as a last resort. (And it keeps the theoretical cylindrical perimeter.) You better know what mesh size you need because you are not going to refine this mesh after you get the first set of results!

Figure 1: Create a mesh of a repeating pattern in Inventor. (I should have done the two half holes to make this a "square" pattern.)

Figure 2: Copy the pattern around the perimeter. (I did this using Excel and FNO Reader to translate the calculated nodes and elements to a Nastran file. The finished ring is shown in the Nastran Editor. You will not be able to view the results or create the load in Inventor; it is all done manually/programmatically.)

Figure 3: Ten-sided polygon with 3000 holes total; 73.5k nodes, 70.8k elements. Meshes in less than 1/2 minute. (Started with solid polygon and "drilled" one hole. Arrayed the hole on the flat face. Arrayed the holes around to the other faces. Deleted all the faces except the outside face to leave a surface mesh. In hind sight, I would have done one face only and arrayed the face around the perimeter, assuming Inventor can do that.)

John