Substructure or Submodeling 1

[Lee.Conti]

Hi, I'm running a series of dynamics analysis (modal, random and transient). I wanted to do a submodeling to the assembly model for a small area of interest instead of doing local meshing refinement.

However, I understood that submodeling seems not appropriate for dynamics analysis as the displacement and loads vary at different frequency. Therefore, substructure is recommended... Any thoughts?

Thank you!

 

[FEA way]

Which software do you use ? I will tell you how it looks like in Abaqus as it's probably quite similar everywhere. In this program out of all frequency domain procedures submodeling can only be used in direct solution frequency response analyses. What's more, maximum and minimum frequencies specified for submodel simulation must lie within the range of frequencies calculated in the global model for the results to be accurate. Solution will be the most accurate around the natural frequencies of the global model. On the other hand substructures can be used in almost all dynamic analysis procedures (with a limitation for response spectrum and random response simulations in which you can't calculate results within the substructure). However substructures in dynamic procedures introduce some additional approximations - reduction methods are introduced to reduce mass and damping matrix of substructure. Either Guyan reduction or dynamic mode addition is used for this purpose. The second approach gives much more accurate results since it accounts for dynamic modes of substructure (at a cost of additional eigenfrequency extraction analysis prior to substructure generation).

As you can see substructures are superior in dynamic analysis. However keep in mind that they serve different purpose. Submodeling lets you focus on a small part of a global model, refine mesh in this selected region, optionally also modify its geometry and then use solution from global model to drive this more detailed analysis (without the need to run the whole global model simulation again). Substructures are used to replace repetitive parts of the model or parts of large assemblies with already generated matrices.

 

[Lee.Conti]

Hi FEA way, I am using ANSYS.

 

[FEA way]

I've never used these techniques in ANSYS but general rules are the same and from what I know ANSYS has similar capabilites to Abaqus in terms of submodeling and substructures. So you can perform both kinds of analyses for selected dynamic problems but their applications are different. ANSYS also offers two reduction techniques for substructures - Guyan reduction and component mode synthesis which is generally equivalent to dynamic mode addition.

 

[Lee.Conti]

Are you saying both submodeling and substructuring are fine to use in dynamics condition? I never use superelement before but submodeling in static so want to get some information about dynamics analysis.

 

[FEA way]

Yes, both techniques can be successfully used in dynamic analyses. You just have to be aware of their applications, limitations (mainly in terms of which dynamic procedures support them) and approximations. I described them briefly in my previous post but it will be best if you refer to ANSYS documentation too since there are some software-specific issues.

 

[Lee.Conti]

I found a very interesting comment from ANSYS forum:


"Submodeling in dynamic is not easy because displacement is not enough: we need frequency or acceleration at each node for inertia effect as well as phase information.

In PSD, It is not possible to retrieve 1-sigma displacement because they have no associated frequency. The composition of the modal contributions must be redone "by hand"

In harmonic, it is a little simpler but not standard because it is necessary for each frequency to recover the real and imaginary part of the displacements.

Julien Ravoux"

 

[scheah]

Hi Lee.Conti,

It sounds like submodeling a dynamic analysis in Ansys is what you are attempting. This blog post may be helpful:
Link


Best,
Jason

 

[Lee.Conti]

Hi Jason,

Thank you for your sharing. I came across with submodeling in modal analysis but some commented that it is not recommended in random vibration or other dynamic analysis due to the statistic displacement, perhaps not ideal for random vibration? But I am not sure if anyone has been using it in Random Vibration or implicit dynamics... and that is the reason I am considering to use substructuring instead of submodeling.

 

[scheah]

Hi Lee.Conti,

Fundamentally speaking, random vibration and harmonic analysis both uses the same equations of motion. Where they differ is how the modes are scaled and added as well as how the stresses from different modes are 'added'. But yes, I agree with you. Smartly substructuring should solve your problem.


Best,
Jason