Load scale factor peak

[Anonymous]

Hi there,

I am currently working on a screw connection that is supposed to connect two solid beams. Because there are also transverse forces to be transferred, I have chosen a friction contact between the solids. The connection is affected by transverse forces and bending moments.

Setup (non-linear statics):
Tref: 0K (ALL)

Subcase 1 (5 increments): The screws (beam elements) are "thermally" preloaded (-100Kelvin  -->10KN preload on bolt). Works very well!
Subcase 2 (15 increments): The load is initiated here. "thermal preload" remains.

The model is symmetrical in x and z direction and very stable. Solution converges fast.

At first sight the results look ok, but there are two inconsistencies. It looks like the solver applies the load fully and then ramps down incrementally, see the "Applied Load vs Load Scale Factor" - plot below. The result can also be seen when looking at the "Beam axial forces". The displacement of the force application point seems to be ok.
Even if I remove the external load and only want to keep the preload in subcase 2 the results are almost identical:
Subcase 1: Thermal Load
Subcase 2: Thermal Load (same as Sub1 - to "hold" preload).
The plot looks almost identical...

Applied Load vs Load scale Factor (sub1: preload sub2: preload)

Beam axial Force vs Load scale Factor (sub1: preload sub2: preload)

Stettings:
Nastran in CAD 2018
-SLINESLIDETYPE: DYNAMIC and STATIC tested. Static gives bad results. (Quick question: When i expect large sliding, should i use dynamic?)
-WARNINGS: OFF because of the "TRef message: E2..." [i dont know]

I wonder what the problem is.
-Do the results of the intermediate increments have no significance here?
-Is the contact correct defined?
-Why does the "Applied Load vs Load Scale Factor" - plot have a peak.

Does anyone have an idea?

Thank you!

Alex

[Anonymous]

Edit: Model attached

[John_Holtz]

Hi Alex,

I reproduced your results with version 2018. I am not sure what causes the problem.

I did notice these things. (Load=1.0 is at the end of the preload subcase, and Load=1.2 is the first result in the second subcase. I may have changed the number of steps, and changed the live load to a near-zero value to see how the thermal preload was behaving.):

1. The contact force does not change from load 1.0 to 1.2.
2. The solid von Mises stress does not change from load 1.0 to 1.2.
3. The beam axial stress does change.
4. The displacement does not change from load 1.0 to 1.2.

These factors do not agree with each other. If the bolt stress really increases, I would expect the other results to change too. I suspect the output of the beam stress is incorrect but otherwise the results are okay.

I also ran the model with version 2021. It shows the expected behavior for the beam stress. It is possible that something was fixed between 2018 and 2021.

[Anonymous]

Hi John,

Thank you for the fast response!

I also have an 2021 Version. I ran the model and got the same results as with 2018.

Can you upload your model, so i can work with the correct results an compare this?

The workaround with the "termal preload" is okay, isnt it?

Do you have an alternativ solution for me? 

Are you going to keep an eye on the problem?

Thank you!

[John_Holtz · Accepted solution]

Hi,

I have attached my model in "version 2021.zip". Although the assembly file has a different name, the part files are still the original name. I suggest that you save this file and unzip it to a new folder.

In addition to the change to the live load, you will see that I made a change to the temperature load. Based on this article, I thought the problem could have been the same temperature load in two subcases. (Note that the behavior in your model is different than the behavior in the article, so it was a just a test to see if the problems were related.) See Expansion wrong when using multiple subcases in Nastran.

It is strange that your 2021 analysis gave the same result as 2018. Note that I used Inventor Nastran version 2021.2 which was released within the last 2 weeks. What are the chances that the "fix" was released in that version? Another test that I did was to use the Nastran file (the .nas file) created by version 2021, and I analyzed it using version 2018. I was thinking that one possibility is that something was wrong in the .nas file itself. By my 2021.2 Nastran file gave correct results in the 2021 solver and wrong results in the 2018 solver. So this confirms that the difference is in the solver and not in the Inventor interface.

Let me know what you find out.

[Anonymous · Accepted solution]

Hi @John_Holtz ,

i have tested your "fixed" model and some other models in 2021.2. The "applied-load vs Load factor scale" - plot is still wrong, but the results are as expected and seems right.

Thank you!

Greets