Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 596
Reply with quote  #1 
I've had some frustrating run-ins recently with people misusing modal analysis results. I'm sure many of you have used FEA to estimate the natural frequencies of a system at some point or another. The outputs give you natural frequencies, mode shapes and some form of the mass participation ratio (or the very similar modal participation factor, effective modal mass, effective mass ratio, etc.). Have you used these MPR values in your analysis/interpretation? If so, how?

I'm not looking for the correct answer. I want to know what your typical treatment of these values would be, even if it is to just ignore them.
Walt Strong

Sr. Member
Registered:
Posts: 873
Reply with quote  #2 
https://www.bing.com/search?q=mass+participation+ratio&FORM=ANCMS9&PC=U531

Walt
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 596
Reply with quote  #3 
I know what they are Walt. That was not the question.
Walt Strong

Sr. Member
Registered:
Posts: 873
Reply with quote  #4 
No insult intended! I stay on the experimental and testing side of things and let others do modeling when needed, so my extremely limited knowledge of FEA is evident.

Walt
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 596
Reply with quote  #5 
I was looking for what the prevailing attitude was of people who weren't necessarily familiar with it. After all, it is not the simplest thing to explain in a few lines.

I had a customer give me a modal analysis  and say that all modes with less than 0.5% participation in any direction are irrelevant and cannot be excited by the pump. That is absolutely false. There was a mode with 1% separation margin that they disregarded because it had essentially 0% mass participation factors. I think its a very common misconception.
electricpete

Sr. Member
Registered:
Posts: 634
Reply with quote  #6 
I have a vague memory of the term "modal participation factor" from years back when I was studying modal analysis, and based on that memory I can see some logic in the customer's view. 

I thought this terminology applies to the scenario of a sinusoidal forced excitation (of a given frequency at a given location and orientation). The total forced sinusoidal response is a sum of the forced responses of the individual modes. The modal participation factor describes the extent to which each mode is excited by that excitation (or the extent to which is mode participates in the total response... hence the name).  We could also deduce similar info to modal participation factor by examining factors like how close is the modal frequency to the exciting frequency and how close is the modeshape anti-node to the location where excitation is applied.   

Based on that understanding above (right or wrong),  I would have agreed with the customer in the respect that I would think 0% modal participation factor means that mode is not excited and does not contribute to the forced response (assuming the excitation location and direction  is correctly modeled... which is worthy of questioning since 1x excitation can be generated in diverse ways ... varying unbalance location, misalgnment etc).   If I saw the modal frequency was close to exciting frequency, that would make me step back to figure out what was going on since in my view a mode whose frequency was close to excitation should have a higher participation factor unlesse the assumed excitation was applied at a node of the mode or maybe in an orthogonal direction (if the directional systems are somewhat uncoupled). 

I have so far resisted the urge to go back and review modal analysis and terminology for this thread (so maybe I can better contribute toward representing common misunderstandings).
OLi

Sr. Member
Registered:
Posts: 1,874
Reply with quote  #7 
Yes, I would say that reality rule, if the model is not accepting the reality I know what is wrong.... Eons ago when models were new during design a model that was "perfect" could be used to say design a turbine and you added the normal 30% margin to exciting freq's say 1xRPM w/o any even crude calibration of the model and guess what happened? When in this basic example from the real world long time ago, the bearing properties were way off from the model assumptions? Yes it was a 100% hit btw. excitation and resonance IRL, to bad, some extra work needed.
So in my book if you don't verify your model at least once or even in scale or by know how, you may be very far out whatever your model indicate but that is only my very practical view.
If I in my daily work with the crude tools I have or the perfect tolls a customer have find a model that is pretty close to what is found in reality just in frequency perspective I am happy and may work to change that in the model and apply in reality and it works better than welding and sledge hammer freestyling.
I am surprised as in later decades working with model theory people improved dramatically so I rarely see people disregarding reality and sticking to the perfect theory of the in practice
one way or another less perfect model. So I agree with Epete if reality do not coincide with theory, reality win in my world. 
I have the in the last 10 years seen people fiddling with rotor dynamics like creating a double resonance using 2 bearings and by that get a lower more damped resonance with twin peaks, very nice in theory with all those VFD's and broad operating speed ranges etc. worked a bit so-so IRL, would not really suggest that as a solution but gave us a bundle of work so who is complaining......

__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 596
Reply with quote  #8 
Thanks Pete, I think that is a very easy idea to fall into. Of course, the mass participation factor comes from the modal analysis, which has no excitation associated with it (unlike force response analysis).

Olov, we model pretty much every pumpset we sell over 200kW. The modal analysis of the structure is pretty reproducible. Being 30% is inexcusable. As long as we have accurate information going in, then being within 5% is typical.

Of course, now that I got laid off, I shouldn't be speaking in present tense... Also, this questions matters a lot less, I guess, but I couldn't delete the thread!
OLi

Sr. Member
Registered:
Posts: 1,874
Reply with quote  #9 
It was way back and it was not so accepted either, after the happening. So just have it clearer for me, this was customer input on "your" data? I sure expect you to be within 5% on your own machine these days. Maybe you can do some work for the customer then, he may have something that need adjusting? All the best.
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
electricpete

Sr. Member
Registered:
Posts: 634
Reply with quote  #10 
Ouch, did you just get laid off?  Sorry to hear that. 

I will mention, forced response can be computed from the modal analysis... that was the context I was thinking about.  
  • Modal analysis is not REQUIRED to solve a steady state sinusoidal forced vibration problem since if we have the M and K system matrices we can solve directly the matrix equation X = F / (K-w^2*M)...
  • But if you have the modal analysis completed (eigenvalue problem solved) that provides a transformation from a coupled N-DOF system to N uncoupled SDOF systems (each SDOF system represents a mode).     The total sinusoidal steady state forced output response can be computed as a linear combination of the responses of the N uncoupled systems (with a linear transformation applied on the input excitation also).   It is a much more complicated way to solve the forced vibration problem than the simple X=F/(K-w^2*M), but it gives some insight as to how each mode "contributes" to the forced response.     Mechanical Vibrations by Rao 3rd Ed. section 6.14 (and example 6.17) shows this approach.   They do mention a "modal participation factor" in their discussion of this approach,  but it's a bit different than I had described it (they use "modal participation factor" interchangeably with "principle coordinate", which confuses me).  



Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 596
Reply with quote  #11 

Quote:
Originally Posted by olov
It was way back and it was not so accepted either, after the happening. So just have it clearer for me, this was customer input on "your" data? I sure expect you to be within 5% on your own machine these days. Maybe you can do some work for the customer then, he may have something that need adjusting? All the best.


In a nutshell:
The customer performs a modal analysis on the elevated foundation. They say it passes. Pump goes in and we have vibration despite our modal analysis on the pump itself suggesting separation margins being 10%+. Then we request their analysis report and see that indeed, it should not have been passed, because they misunderstood that 0% modal participation factor does NOT mean insignificant.

Quote:
Originally Posted by electricpete
Ouch, did you just get laid off?  Sorry to hear that. 

I will mention, forced response can be computed from the modal analysis... that was the context I was thinking about... they use "modal participation factor" interchangeably with "principle coordinate", which confuses me).  


Yeah, yesterday, which is why I care significantly less about this problem now. Probably shouldn't even be talking about the specifics.

The forced response analysis I did in my first job was all on the principal of modal superposition, where you run the modal analysis and essentially 'synthesize' FRFs that are then used to very quickly calculate the output from a sinusoidal input. The modal participation factors were an outcome of the modal analysis, and were unchanged by the excitation you chose. Of course, there are many similar dimensionless parameters that are very similar to modal participation factor, and will, some of which would come only from the force response analysis. I imagine Principal Coordinate is just a tweak of the MPF, and they use them interchangeably because they use the MPF only as an indication of direction. Kinda stupid though.

The forced response analysis we did at my recent job were direct time-based analysis, which as you said, precludes the need for a modal analysis to be performed explicitly. Therefore mass participation factors are never calculated. This is required for non-periodic analysis, clearly.
OLi

Sr. Member
Registered:
Posts: 1,874
Reply with quote  #12 
Ah, the end user make his foundation, that is sometimes interesting in part that they or their consultants maybe don't do that so often or the consultant is famous for his distribution pump foundations for heating water distribution or so. Most severe I seen so far was like a 150 T machine train on a 13m hi concrete foundation where they suddenly during the concrete production entered a 0.5 m rubber layer for earthquake isolation w/o recalculating the rest. Defining that it would not make any difference.... It did and explained the strange behavior and looks of the OEM testbed design at FAT that I could not understand when I saw it and OEM just said it was dynamically as the reported design on site. It did not follow the design rules of the machine part OEM that I worked for anyway. So a resonance is a resonance and if it does give problems no math magic can change it more than supplying a tool to solve the problem in the best way so it usually get solved one way or another. Take care and I hope the world can be reworked to a more normal state in reasonable time. I am not so sure but I hope.
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.