Response Spectrum Results in Ansys

[ShakeNBake444]

I'm using workbench in version 15 and I have a question about response spectrum results. Simply put, what in the world do the results mean for displacement? The reason I ask is simple: the input required is some input frequency spectrum. Lets say for the sake of argument, I have a simply supported beam and at 10 Hz my input is 2 G's and 20 Hz my input is 1 G. That is the input spectrum. When i run it and request a displacement output, i get only one plot with displacement. I would expect a plot for each frequency bin. Obviously this would be a huge file if you use more than a few frequency bins. So then the question is, well which response is it? is it the response from 10 Hz or 20 Hz or maximum of the two or something different? Naturally i cannot find anything from the Ansys documentation that explains what is being plotted with this output is requested. Any thoughts?

 

[Drej]

I think you're pushing buttons with your eyes closed. Have you bothered to take the time to understand what you're doing before doing it?

Response Spectrum Analysis (RSA) isn't like doing a simple static FE analysis, it is actually quite involved, and can require a lot of interpretation. Take the time to talk to someone about this, do some background reading (read the ANSYS manual - I think you'll find it's excellent if you take the time to go through it), and then run a simple benchmark to understand what RSA is all about.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[rickfischer51]

Response Spectrum Analysis is a giant idealization. It starts by running a modal analysis. It takes every modal frequency within the range your response spectrum is defined, and finds the g-loading at that frequency. Then it calculates mode participation factors, mode coefficients,etc, which are used to scale those g-loads into realistic loads. It throws out insignificant modes, then applies each in a linear elastic structural solution. Finally, it combines those results using a mode combination method to produce a single result.

DREJ is right. You should start slow and understand the process before you plunge in. And use MAPDL to learn, not Workbench. MAPDL will be more transparent and make it easier to understand what goes into the process. ALso, use the MAPDL documentation. Almost everything I know about spectrum analysis I learned by reading the MAPDL documentation. Workbench documentation sucks.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[ShakeNBake444]

I look through the documentation for "MAPDL" and nothing telling me what that actually is. I'd love to learn about these things buy Ansys doesn't make it easy. Is MAPDL just a different flavor of APDL? Rick, your answer is much more helpful than the first, but please understand that answering people's question with "you should know more about what you're doing first" is not helpful at all. After all, that's why we are here in the first place. I think i know enough about it to at least make a first attempt. We don't know everything about riding bikes before we get on one for the first time.

The documentation for response spectrum says the following:

The results from the ANSYS solver are displayed as the model’s contour plot. The results are in terms of the maximum response.

Ok, that's clear. And I believe that answers my question.

Let's simplify my initial question a bit for the sake of argument and I will present what I think ansys is doing with my simple model and you can tell me if I'm wrong or otherwise. Imagine a one degree of freedom system: a spring and a mass. Spring constant is 10 N/m and the mass is 5kg. So the system has a natural frequency wn = 0.225 Hz. I apply a base excitation of 4*wn. I can calculate the response easily (assuming no damping) with a formula for the coefficient for the particular solution, which is wn*Y*sqrt(wn^2/(wn^2-4*wn^2)^2). Where Y is 1 meter peak to peak excitation. I would see a 0.333*cos(blah blah blah). So if it took this same system into ansys (assuming it's possible) in to a response spectrum analysis and I asked to look at the frequency of 4*wn with an displacement excitation of 1 meter, would I see 0.333 in my displacement plot? I think the answer is yes because of what i just layed out. Is this true?

Thanks

 

[rickfischer51]

MAPDL is Mechanical APDL, aka Mechanical, Ansys Classic, Ansys blackscreen, Ansys. It is the original Ansys Program. The full program is Ansys Multiphysics, then Ansys Mechanical which is just structural and heat transfer, then Structural which is Mechanical minus heat transfer, etc. Workbench is a front end that writes a script and submits it to MAPDL. APDL is the Ansys Parametric Design Language that drives MAPDL. Or thats how it used to be before R17. I guess they needed the Mechanical name for the original Workbench offering as they morphed the Workbench moniker into an environment. So the original structural and thermal Workbench became Mechanical, Mechanical became MAPDL, etc.

Look in the help table of contents, Mechanical APDL, Structural Analysis Guide, Chapter 6.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[Shu Jiang PEng SE]

To my understanding. Each modal of a structure will have different response to given excitation with specific frequency. The maximum response of each modal of the structure generally does not occurs at the same time.


Response spectrum analysis perform the following operations:

1. Get the maximum response of each modal of the structure under a given excitation with specific frequency
2. Combine the maximum response of each modal per designated method, therefor get the theoretical maximum structure response under given excitation
3. Repeat step 1 and 2 to get structure maximum theoretical responses under excitation with different frequencies

 

[rickfischer51]

"2. Combine the maximum response of each modal per designated method, therefor get the theoretical maximum structure response under given excitation."

I don't think this is strictly correct. It typically does not apply the maximum response. The response at each modal frequency is scaled via participation factors, mode coefficients, etc into realistic loads, a significance factor is used to remove insignificant loads, and then the expansion pass is made with these remaining reduced load. These results are combined with a method like SRSS, CQC, DSUM, etc, where they are further reduced based on their proximity to the dominant frequency.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[Shu Jiang PEng SE]

Yes. I agree with rickfischer51.

The final results from spectrum analysis not the real ones. The reasons are: 1. The maximum response of each modal do not occur at the same time, some technical combination methods are used to "guess" the possible results; 2. some high frequencies modal responses are neglected in combination.

 

[ShakeNBake444]

"The final results from spectrum analysis not the real ones"

Then why would you do this in the first place if in the end you get a result that is not real at all and does not represent the system under the given excitation?

 

[rickfischer51]

You do this because it is conservative. You have two other options. Take the highest g loading anywhere in your spectrum and apply that load. Or run it as a transient dynamic and spend half of what remains of your life waiting for a solution, and then spend the other half postprocessing. The first is often prohibitively over conservative, the other is an analytical nightmare.

Actually there is a third option. IF you examine the mode combination methods you will see that if modei is the dominant mode and the frequency of modej is sufficiently far from the frequency of modei, it can be ignored. Sufficiently far is about 10%. Then check your modal mass and use the missing mass technique to compensate for the modes you are not including. This gives you an equivalent static load that you can now apply in a single run. An advantage of this technique is that you can now apply this to a nonlinear analysis. Spectrum analysis relies on modal superposition, which is only valid for linear results. A disadvantage is I don't think this is specifically supported by any of the codes. I think ASCE 7 allows you to derive an equivalent static load however you want, as long as you can justify it.

Spectrum analysis is an idealization of an idealization. It is a static idealization of a dynamic event applied to a linear idealization of a nonlinear structure.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[Shu Jiang PEng SE]

Two purposes of spectrum analysis: 1. Evaluation of the structure responses under excitation with different frequencies; 2. Most important, tell the excitation frequencies the structure is "sensitive" to. Therefore design the structure lower frequencies away from the possible excitation frequencies.