Abaqus FSI methods: CEL vs CFD-Standard/Explicit co-simulation vs? 1

[biofriendly]

Hi All,

I am new to considering dynamics of fluids in a model. I am learning of the ways Abaqus can handle fluid-solid interaction. By fluid I definitely dont mean hydrostatic as that can be handled in Abaqus/Standard.

I already have practiced a few examples in Eulerian-Lagrangian domain using Abaqus/Explicit. I never have used CFD so no idea. Any comments much appreciated!

I was wondering what are the major differences between 1-Eulerian-Lagrangian and 2-co-simulation coupling Abaqus CFD and either of Explicit or Standard modules? and when to use which of them?

Plus, does Abaqus offer any other additional techniques other than Coupled Eulerian or CFD to handle fluid-solid interaction problems?


Thanks a lot!!

A

 

[StefCon]

Hello, I am trying to learn this myself. I got the impression that co-simulation is more for static boundaries. There is also SPH (smoothed particle hydrodynamics (?)) which I've seen been used for bird impacting turbine blades. I think the SPH is particularly good for very violent flows. I also got the impression that CEL is easiest to get working (not perfected yet...).

This is a very interesting topic. I do not want to hijack your thread but you might find it interesting. I made an example model using CEL but the results seem strange.

Link

Has anyone in the forum any hands-on experience with this type of analysis?

/S

 

[IceBreakerSours]

That is a lot of questions! To be honest, these are advanced questions and, in order to capture the meaning, you will have to cross a threshold first - which is fun because there is so much more to learn. However, the documentation is an excellent resource to get a few quick answers. That being said, here are my two cents:

Depending on the amount of deformation a given physical phenomenon undergoes, within Abaqus, one may choose anything in between a purely Lagrangian analysis or a purely Eulerian analysis or something in between (Arbitrary Lagrangian-Eulerian with adaptive meshing or Coupled Eulerian-Lagrangian). On the other hand, SPH/DEM fall under the broad category of the so-called meshless methods; the documentation is pretty good at suggesting when to use these methods. In the case of DEM, it's the particle-particle interaction that one cares more about. On the other hand, SPH is a better tool to use when fluid-solid interaction is of interest.

Abaqus employs either an implicit method (Abaqus/Standard) or an explicit method (Abaqus/Explicit) to solve the physics (i.e., differential equation converted into a discretized form). Both solvers, theoretically speaking, MUST give the EXACT SAME results. In many problems, that happens to be accurate. However, certain practical limitations force one to choose one or the other solver. The terminology comes from linear algebra and, if you have no idea, it will be immensely helpful to grab a good book (Strang, for example) and go through it. A co-simulation is helpful when you wish to use both the implicit and the explicit solvers in tandem.

Bottomline: Which solver/analysis type .. .. you choose is primarily governed by the physical phenomenon that you wish to capture. Each of these will come with its trade-offs. For example, CEL is great at capturing huge deformation and FSI, but it comes at an enormous computational expense. For instance, a question that a typical analyst might struggle with on a daily basis is: "Do I have the hardware/license tokens/time/support/blessings from management .. to accomplish what I must in a week/month/quarter?"

Here's another one that you might hear about: Isogeometric analysis. Abaqus has not implemented it (although the most recent versions of general contact seem to have borrowed some ideas) but LS-DYNA and FEAP have implemented these features. There is free plug-in for Abaqus. In IgA, the basis functions are the same as in CAD, whereas in FEM, basis functions are polynomials. While that may sound like a trivial difference, the choice of NURBS/B-Splines/T-Splines as basis functions has brought a revolution in computational mechanics (mostly in the academic world, so far).

Are you new to this forum? If so, please read these FAQ:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[biofriendly]

Thanks IceBreakerSours and StefCon for your replies and contribution to understand these topics for me. Although the answers actually brought me with more questions than answers. I will also dig in more and will update here of my findings...

 

[IceBreakerSours]

While the Abaqus documentation is probably the best in the business, you should also google some of your basic querries; odds are someone somewhere asked the same questions and received excellent replies.

Are you new to this forum? If so, please read these FAQ:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[DrBwts]

Great post IceBreakerSours.

I have only used fluid solid interaction (FSI) in Adina & LSDYNA so I dont know the specific ins & outs of the implementation in Abaqus. It is very important that you use the right modelling strategy. If the dynamics & transients in you model are important, or you have high deformation then explicit is the way to go in my opinion. As far as the FSI goes, if you have large deformations under high impact (shock wavers etc) then LSDYNA is the way to go.

 

[IceBreakerSours]

While I agree with your outlook, DrBwts, I offer a different take

Someone not too long ago complained to me that while their LS-DYNA model was working fine, the same model in Abaqus wouldn't work. [Several years ago I learned a lesson: If you think these codes are wrong, think over and over again.] I picked up DYNA and found out there was a flag in LS-DYNA (AUTO SPC or something) that was allowing the model to run by adding stiffnesses automatically (not a great idea, by the way) and they had never bothered to even look at the artificial energy. Fortunately, it didn't matter since the energy was sufficiently low but the story illustrates the point that I wish to make: You can't go wrong with any one of the major codes; it is the pilot on whose shoulders the model is standing.

That said, each major code seems to have its strengths and, as you said, DYNA is famous for explicit analyses (explosions - which was its first major application area, crashes, etc.), ANSYS seems to have gained a stronghold in multiphysics, whereas Abaqus seems to be good at materials. [Some of this happens because of key business decisions, such as acquisitions.] However, in my biased opinion, no code has nearly as good a documentation as Abaqus. For instance, to be honest, I find the element/material .. naming convention in DYNA awkward.

Are you new to this forum? If so, please read these FAQ:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[DrBwts]

I agree Abaqus has by far the best documentation out there & the Python API is fantastic.