NEi Nastran vs NX Nastran

[gg8014]

Recently our company has transitioned from using NEi Nastran to NX Nastran as our FE solver. In doing so we notice some significant differences in the results for the same models. Our models are fairly straightforward using linear static SOL101. One of the first differences that we noticed was the displacements for the RBE3 elements used to connect point masses to our structures. The displacements using NEi were small in comparison to the NX results especially as the mass values increased. We posed this question to our support representative. His response was to remove the rotational degrees of freedom from the RBE3s independent nodes. We did this and the large deflections decreased such that they were comparable (although not exactly the same) as those when the model was run in NEi. On comparing the results of the constraint forces and moments between the two solutions we saw that there were differences in some cases approaching 10%. Again we are using the exact same model in both solutions. While this might not be an issue for some situations it is an issue when models that were accepted by our clients are re-run using the NX platform (at their insistence) and they see that the results have changed. The question is which platform is producing the more correct results?

 

[JBlack68]

Difficult to say what's the source of the problem is without looking at the model(s). But...
1/ is the sort of testing the company should have carried out way before doing the transition as part of a de-risking exercise
2/ does the problem exist for every single model tested?
3/ check the parameters default values (k6rot, snorm, etc).
4/ in my experience (and in my part of the world) the NX support is the worst I have seen over nearly 20 year so best of luck
5/ go back to a very very simple model. A plate with a mass attached via an RBE3. I can only think of the way the RBE3 is defined ie dof transmitted
6/can only strongly recommend to push the problem on the supplier. (Mind you it depends on the T&Cs negotiated as part of switch)

 

[ThomasH]

Hi

I remember when we went from MSC to NEi, there were some questions regarding the solvers. I our case it was primarily the speed of the solvers.
In the end the answer was that the settings we used that made MSC fast were not ideal for NEi. So when we changed some parameters the solution times were similar.

Regarding differences in the results I can imagine that there can be settings that will change things on a detailed level. Since I have had many discussions with the Tech's from NEi I'm fairly certaion that if the problem was reverse (going from NX to NEi) they would help you. But since you left NEi it's not the same thing.

There are settigs in the editor interface that controls certain parts of the analysis. To get the exact same settings in NX should be possible but I don't know how. I know there are tricks in MSC and I suppose they exist in NX since they they have the same origin. But comparing them on that level, sorry, it's been a while since I used MSC on any level.

But I agree with JXB0809, take it to the supplier.

Best of Luck

Thomas