tank not stable with rotation constraint applied

Discussion created by Thomas Bayley on Jul 7, 2020
Latest reply on Jul 14, 2020 by Thomas Bayley

Like • Show 0 Likes0
Comment • 6

Hi,

I have a round tank model with bottom reinforced structural underfloor support welded onto the outside surface. The model is at the assembly level with two parts (tank and support). The tank is fixed at the bottom edge of the strake. The tank is constructed with the surface for FEA purpose (real model is not surface), bottom support is with real weldment. It is global bonded and all the relevant constraints have been applied. After evaluating the model stability, the program told me that one rotation constraint about the center axis is missing on the group of support and bottom floor. I have applied a rotation constraint on it, but it brings more missing constraints instead of fixing it. I am currently confused that why introducing one more constraint can make the model more unstable?

Can anyone please instruct? Or how to fix this issue. The model is attached for reference.

Attachments

Outcomes

Attilio Colangelo Jul 8, 2020 1:35 PM

Short answer - Trial and error

Long answer

It looks like the problem is with the part "Tank -FEA.sldprt". It is having a hard time attaching the sloped bottom to the tank walls. I had to knit all the contiguous vertical surfaces of the tank wall together. I was able to solve it by explicitly adding a bonded contact set between the edges of the sloped bottom and the faces of the touching cylinder. And for some reason you need to remove global contacts.

The feature "Boundary-Surface1" may not be necessary for the analysis butting the sloped edges directly onto the cylinder would be cleaner.

You need to do the same for the assembly although the global bonded seems to be ok there. You can remove the mesh controls. They don't seem to be helping.

Hope that helps.

Actions

Thomas Bayley @ Attilio Colangelo on Jul 9, 2020 8:37 PM

Hi Attilio,

Thanks for your reply.

Yes, bonding between the floor and cylinder can get the model stable. But I wonder why global contact doesn't bond them together?

I didn't delete the global contact by adding a new local contact, and the model can be solved.

The reason why adding knuckles to the floor is this area is a critical area experiencing the most deformation. So it can't be deleted from the simulation.

I have deleted the mesh controls as you suggested to get a similar result, but not the one on the knuckle.

Actions

Attilio Colangelo @ Thomas Bayley on Jul 10, 2020 7:44 AM

Knuckle.....that's the word I was looking for.

Global bonding seems to work better when joining different parts in an assembly. The floor plate and the tank are the same part. Another issue is the tangency between the knuckle and the sidewall, the algorithm might just be "missing" that juncture.

On a separate note if the floor plate is an area of concern you may want to consider modeling the beams as solids. Pure beams may not give the support (mathematically) that you would actually be getting so it would be a good comparison.

Actions

Thomas Bayley @ Attilio Colangelo on Jul 12, 2020 8:37 PM

Yes, Attilio, thanks for your reply.

It seems like the global bonding has its limitation.

The knuckle is modelled tangent to the wall. So there shouldn't be any connection problem between them. And they should be bonded in the part level. However, there is a mathematical hardship within the software itself to not be able to calculate properly on the connection.

This is my understanding. Please correct me if I am wrong.

I change the bottom support to solid, and the result is much better and realistic. I was concerning the calculation time when applying beam.

Actions

Attilio Colangelo @ Thomas Bayley on Jul 14, 2020 10:51 AM

Yes, the "automatic" contact generation is better when there is a sharp (90 degrees?) juncture. Sometimes the software needs help figuring that out which is why manual contact sets need to be defined sometime.

That's good news on the solution for the solids. I think that gives you a good appreciation for what's happening "under the hood" with the different element types.

Actions

Thomas Bayley @ Attilio Colangelo on Jul 14, 2020 8:16 PM

Thanks so much, Attilio.

It is good to know the limitation of global contact on part and assembly level.

It seems like the SW tries to simplify the meshing by using beam instead of solid, but it is not working very well. Especially two of the beams in the model were forced to mesh with solid.

Anyway, it's all clear now.

Actions

6 Replies

Attilio Colangelo Jul 8, 2020 1:35 PM
Short answer - Trial and error

Long answer
It looks like the problem is with the part "Tank -FEA.sldprt". It is having a hard time attaching the sloped bottom to the tank walls. I had to knit all the contiguous vertical surfaces of the tank wall together. I was able to solve it by explicitly adding a bonded contact set between the edges of the sloped bottom and the faces of the touching cylinder. And for some reason you need to remove global contacts.
The feature "Boundary-Surface1" may not be necessary for the analysis butting the sloped edges directly onto the cylinder would be cleaner.

You need to do the same for the assembly although the global bonded seems to be ok there. You can remove the mesh controls. They don't seem to be helping.

Hope that helps.
Like • Show 1 Like1
Actions
- Thomas Bayley @ Attilio Colangelo on Jul 9, 2020 8:37 PM
  Hi Attilio,
  Thanks for your reply.
  Yes, bonding between the floor and cylinder can get the model stable. But I wonder why global contact doesn't bond them together?
  I didn't delete the global contact by adding a new local contact, and the model can be solved.
  The reason why adding knuckles to the floor is this area is a critical area experiencing the most deformation. So it can't be deleted from the simulation.
  I have deleted the mesh controls as you suggested to get a similar result, but not the one on the knuckle.
  Like • Show 0 Likes0
  Actions
  - Attilio Colangelo @ Thomas Bayley on Jul 10, 2020 7:44 AM
    Knuckle.....that's the word I was looking for.
    Global bonding seems to work better when joining different parts in an assembly. The floor plate and the tank are the same part. Another issue is the tangency between the knuckle and the sidewall, the algorithm might just be "missing" that juncture.
    
    On a separate note if the floor plate is an area of concern you may want to consider modeling the beams as solids. Pure beams may not give the support (mathematically) that you would actually be getting so it would be a good comparison.
    Like • Show 0 Likes0
    Actions
    - Thomas Bayley @ Attilio Colangelo on Jul 12, 2020 8:37 PM
      Yes, Attilio, thanks for your reply.
      It seems like the global bonding has its limitation.
      The knuckle is modelled tangent to the wall. So there shouldn't be any connection problem between them. And they should be bonded in the part level. However, there is a mathematical hardship within the software itself to not be able to calculate properly on the connection.
      This is my understanding. Please correct me if I am wrong.
      I change the bottom support to solid, and the result is much better and realistic. I was concerning the calculation time when applying beam.
      Like • Show 0 Likes0
      Actions
      - Attilio Colangelo @ Thomas Bayley on Jul 14, 2020 10:51 AM
        Yes, the "automatic" contact generation is better when there is a sharp (90 degrees?) juncture. Sometimes the software needs help figuring that out which is why manual contact sets need to be defined sometime.
        That's good news on the solution for the solids. I think that gives you a good appreciation for what's happening "under the hood" with the different element types.
        Like • Show 0 Likes0
        Actions
        Thomas Bayley @ Attilio Colangelo on Jul 14, 2020 8:16 PM
        Thanks so much, Attilio.
        It is good to know the limitation of global contact on part and assembly level.
        It seems like the SW tries to simplify the meshing by using beam instead of solid, but it is not working very well. Especially two of the beams in the model were forced to mesh with solid.
        Anyway, it's all clear now.
        Like • Show 0 Likes0
        Actions

tank not stable with rotation constraint applied

Attachments

Outcomes

Related Content

Recommended Content