Impact Hammer vs. FEA 2

[WARose]

One specialty area I offer as a structural engineer is design to mitigate vibrations from new equipment and I also look at existing vibration issues. (I.e. see if the supporting framing/foundation was designed correctly or if this is simply a situation of machinery getting excessive wear and producing excessive unbalanced forces.)

Regarding the latter.....I get a bunch of situations where I go out to look at something and I get some guy in my ear going (after I tell them I am going to model the situation with the FEA tools I have): Can't ya just whack it with a hammer and get all that?

My reply is just pretty much always: no. My reason(s):

1. If it's on steel....why? There is no mystery there as to the modes. You model it....and you know. Yes a modal hammer could probably give you the same thing with enough whacks. But you'd be beating on it all over the place with that many degrees of freedom....and there would be tons of data to sort.

2. If it's on concrete.....if we are taking a heavy foundation (say 50-300 kips in weight) my experience is: you are likely not going to be able to excite all the modes (especially some of the rotational ones). It might help to establish some soil spring constants. (Always an unknown.) But I can typically come up with them myself. (Or with testing that can be cheaper than modal hammers.)

3. Misinterpreting the data. I've seen some much data from hammer testing misinterpreted and just plain screwed up.

So I am off base to always want to approach it this way? Granted it is typically more expensive.....but to me it's more accurate than the guessing games with modal hammers.

 

[scheah]

I believe modal testing is complementary to FEA.

If you are confident in your FEA model, no modal testing is required. After all, the model can represent present structure with any possible future changes accurately. Modal testing could come in handy to verify the model or help with uncertainties (e.g. unknown inertia values, no CAD or nebulous boundary conditions etc).


Kind regards,
Jason

 

[GregLocock]

My background is in both FEA and experimental modal analysis on vehicles and components.

1 you don't know the boundary conditions, and your bolted joints have interesting effects. FEA doesn't tell you the damping.

2 use a shaker. use more than one if necessary

3 yes.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[WARose]

Thanks (as always) for the feedback Greg.

1 you don't know the boundary conditions, and your bolted joints have interesting effects. FEA doesn't tell you the damping.

Good point on the joints. Although I typically don't have too many questions on the damping for steel. (We are typically talking 2-5%.) In what I do, if you need damping to bail you out, that means you are too close to resonance. Unless we are talking foundation damping which can be large.)

2 use a shaker. use more than one if necessary

I need to look into that one more. Some of the foundations I typically look at have (dynamic) stiffness(s) on the order of 1000-10000 kips/inch. So I don't know if shakers can get any info on that or not. But I guess I need to find out.

 

[scheah]

WARose,

I'm confused. Didn't you say:

There is no mystery there as to the modes. You model it....and you know

With that kind of confidence, would the joints be an issue?


Kind regards,
Jason

 

[WARose]

WARose,

I'm confused. Didn't you say:

(WARose)
There is no mystery there as to the modes. You model it....and you know

With that kind of confidence, would the joints be an issue?

For low frequency stuff, no: I don't wonder too much about the joints. But with the high frequency stuff (when it can become a wave propagation problem; which I have rarely dealt with) Greg is right: some interesting things can happen at the joints. (I.e. reflection, mode conversion, ect.) So my comment about confidence dealt more with low frequency problems......when it comes to high frequency....Greg's point is well taken.

 

[GregLocock]

Incidentally I used to have a Zonic hydraulic shaker which could bend cars. The main problem with it was the load cell, designing a stinger that would transmit the forces without bending the load cell was important.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[WARose]

Incidentally I used to have a Zonic hydraulic shaker which could bend cars. The main problem with it was the load cell, designing a stinger that would transmit the forces without bending the load cell was important.

That's another good point I wonder about with this type of testing: the localized effects (either from the testing instrument or chipping/bending the structure/surface it is in contact with).

 

[271828]

"1. If it's on steel....why? There is no mystery there as to the modes. You model it....and you know."

Have you compared your FE results with measurements? Not just frequencies, but shape and order of modes and frequency response function values.

I'm familiar with the literature regarding modal tests of floors and similar structures. The predictions are often inaccurate. That is my experience, also. I'm going with experimental values, or a combined program of measurements and calculations, if at all possible. I only go by calculations when there's no other option, and I'm up-front with the client that the prediction might be off by quite a bit.

 

[Tmoose]

"...if the supporting framing/foundation was designed correctly or if this is simply a situation of machinery getting excessive wear and producing excessive unbalanced forces.)

In my experience a third, very common cause of excessive vibration at 1,2,3,4 X rotational frequency is the installation is deficient in regards the machine-to-foundation interface. That has been true of new or used equipment, and sometimes in very creative ways. For instance, even if the grout installation looks to be of high quality, evil lurks below the surface.

 

[WARose]

[blue](271828)[/blue]

Have you compared your FE results with measurements? Not just frequencies, but shape and order of modes and frequency response function values.

I'm familiar with the literature regarding modal tests of floors and similar structures. The predictions are often inaccurate. That is my experience, also. I'm going with experimental values, or a combined program of measurements and calculations, if at all possible. I only go by calculations when there's no other option, and I'm up-front with the client that the prediction might be off by quite a bit.

Interesting. In your experience, what kind of floors are we talking about (and what degree of accuracy is needed)? Light framing for clean rooms or other vibration sensitive areas? I can see (in that situation) the dynamics becoming a bit tougher. In part because the non-structural components (possibly not modeled) contributing to the overall stiffness and (of course) the damping becoming even more of a question.

That's a little different than what I typically do.....but I always like to hear about different experiences.

[blue](Tmoose)[/blue]
"...if the supporting framing/foundation was designed correctly or if this is simply a situation of machinery getting excessive wear and producing excessive unbalanced forces.)

In my experience a third, very common cause of excessive vibration at 1,2,3,4 X rotational frequency is the installation is deficient in regards the machine-to-foundation interface. That has been true of new or used equipment, and sometimes in very creative ways. For instance, even if the grout installation looks to be of high quality, evil lurks below the surface.

And that one is about the toughest to spot. (In my experience.)

 

[271828]

Interesting. In your experience, what kind of floors are we talking about (and what degree of accuracy is needed)? Light framing for clean rooms or other vibration sensitive areas? I can see (in that situation) the dynamics becoming a bit tougher. In part because the non-structural components (possibly not modeled) contributing to the overall stiffness and (of course) the damping becoming even more of a question.

That's a little different than what I typically do.....but I always like to hear about different experiences.

Typical steel framed floors and concrete floors. Natural frequencies between about 3 Hz and 15-20 Hz.

It's easy to predict natural frequencies approximately equal to the actual ones. The problem is that the predicted and actual mode shapes usually won't closely match. The modes will have different shapes, be in a different order, there will be a few predicted ones that don't correspond to an actual mode, and there will be actual modes that don't look like any of the predicted ones. Because of this, the effective mass predictions are all over the place. This is probably why most researchers only show frequency comparisons in the papers. FRF predictions look like crap. The problem is that the subsequent calculations are closely tied to the effective masses / FRF mags.

This seems true regardless of whether or not non-structural elements have been installed. The most difficult cases often look like the easiest ones, such as if there are many nominally identical bays.

This is why, whenever possible, I'll go with measurements, or a combined approach with measurements and predictions.

 

[WARose]

Typical steel framed floors and concrete floors. Natural frequencies between about 3 Hz and 15-20 Hz.

It's easy to predict natural frequencies approximately equal to the actual ones. The problem is that the predicted and actual mode shapes usually won't closely match. The modes will have different shapes, be in a different order, there will be a few predicted ones that don't correspond to an actual mode, and there will be actual modes that don't look like any of the predicted ones. Because of this, the effective mass predictions are all over the place.

Interesting. And we are talking frequencies/modes with significant mass participation? With low damping? Significant differences?

Usually I don't wonder too much about structural steel in that situation.....perhaps I should re-think that.

Wonder how much the fact it is impact (i.e. high forcing frequency) affects results.

 

[GregLocock]

It's a similar problem on car bodies and engine blocks. You can get the modal density right (modes per octave), but the agreement of modeshapes in frequency order was pretty nebulous after the first five or so, at least when I was involved 22-35 years ago. I expect that has improved- I'd hope so! The only time I did the modal analysis (at which I am quite good) AND the FEA (at which I am a hack) I got quite good correlation, but had to do many adjustments to my FEA to get that to work. And that was just a 5 kg casting, free-free.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[271828]

Interesting. And we are talking frequencies/modes with significant mass participation? With low damping? Significant differences?

Usually I don't wonder too much about structural steel in that situation.....perhaps I should re-think that.

Wonder how much the fact it is impact (i.e. high forcing frequency) affects results.

Yes, yes, and yes. LOL

Typical damping values are 0.02 to 0.05.

For example, say we have a floor with 10-15 bays and we're trying to predict the response of one bay. We'd compute the natural modes and try to figure out which is responsive in the bay under consideration. The most responsive predicted mode might have motion in 12 of those bays, but in reality, only 6 are in motion. That means we probably over-estimated the effective mass by roughly two, so we underestimated the response by roughly two. There's nothing exotic going on there. It just doesn't seem possible to really nail the modal prediction, and the effective masses are sensitive to these shapes.

 

[WARose]

For example, say we have a floor with 10-15 bays and we're trying to predict the response of one bay. We'd compute the natural modes and try to figure out which is responsive in the bay under consideration. The most responsive predicted mode might have motion in 12 of those bays, but in reality, only 6 are in motion. That means we probably over-estimated the effective mass by roughly two, so we underestimated the response by roughly two. There's nothing exotic going on there. It just doesn't seem possible to really nail the modal prediction, and the effective masses are sensitive to these shapes.

And a modal hammer is capturing that (for 15 bays)? You must really be beating on that thing all over the place. (Or hitting it/shaking it with one heckuva force.)

Thanks for the feedback by the way. (It was me who gave your earlier post a star.) It's rare I get to talk to anyone who does this for structural steel/concrete. I've encountered a lot of skepticism from colleagues on this over the years. (For the reasons I stated in my OP and a few others.)

 

[271828]

We would usually use a floor shaker for those. Heel-drop tests can get a fairly large area moving if the bays are not large and the floor has a non-large psf.

Later. Good luck with your work.