Time Domain vs. Frequency Domain Acoustic Analysis 2

[DLite30]

Can anyone comment on the benefits of performing an acoustic analysis on a recriprocating nat gas compressor, via time domain method versus frequency domain.

I understand that software for performing time domain analysis has only been in use for a couple years, while frequency domain anlaysis goes back 50 or 60 years (maybe more), and has been widely successful.

However, as the end user, a couple of analysis companies are saying that doing studies in the time domain is the greatest new thing...while a couple of the analysis companies that we typically use, have not adopted this method, and still use frequency domain anaylsis.

I have a vague understanding of the fundamental difference in the two approaches, however, I'm unceratin of the differene in accuracy of the two. Does time domain actually predict actual operation more closely...and is this the way that that technology is going? Or, should we just stick to the tried and true method we've relied on for years, and let the technology grow until more people accept that method until we jump into it?

Opinions, please.

 

[GregLocock]

Can you give a link to what you mean by time domain analysis?

Sounds like salestalk.

There are some time related analyses that are worth doing, but I'd always start in the frequency domain.



Cheers

Greg Locock


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

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

 

[hacksaw]

Interesting, is it common practice to use correlation methods prior to performing spectral analysis?

how do you handle the dc component?

 

[Elabbasi]

Both transient and frequency domain acoustic analyses have been around for a long time, and I don't mean just in theory, but in computational software. Is there a specific advancement in acoustic time domain analysis that these softwares claim?

For most acoustic problems the interest is in a steady state solution due to harmonic excitations which is ideal for frequency domain analysis. There are some exceptions like simulation of explosions for example, which are more efficiently performed in the time domain. What is the issue in the simulation of reciprocating natural gas compressors that warrants time domain analysis?

Nagi Elabbasi
Veryst Engineering

 

[drmuh]

Depends on the problem you are working on. You may use frequency domain or time domain analysis. Or you may use both by using wavelet transforms.

http://www.novosim.com

Product Development Services (cad/cae/test)
Medical Device R&D

 

[jeyaselvan]

Probably with reciprocating machines, the pulsation levels may too high to apply linear wave theory. From an academic standpoint, non linear wave equations could be better solved in time domain analysis than in frequency domain. Nevertheless, for all industrial applications, they should not make a lot of difference.

Jeyaselvan

 

[DLite30]

I got my answer, and jeyaselvan is on the right track.

To me, the end-user, basically, is that both time-domain and frequency domain both agree on the frequency response of an reciprocating gas system. However, the difference is in the amplitude prediction accuracy of the two methods...time domain accurately predicts the amplitude levels across the frequency spectrum, while a frequency domain analysis will underestimate the response and not in a consistent manner.

So if one can seperate forcing frequencies and mechanical natural frequencies by a good margin, then frequency domain is sufficient.

However, to more accurately predict shaking forces due to unblanaced acoustic forces, then time-domain analysis is the way to go.

So far, we have had good results with our tech company who solely uses frequency domain analysis. However, not a lot of testing has been done to compare pulsation amplitude levels to the original levels predicted by the model.

Hope this helps some of you in the future.

 

[Hatch]

Just a quick comment.

Time Domain Analysis has been around a lot longer than a few years. It just had a different name but is the same thing.

It was called the "Chart Recorder". You got amplitude versus time. Faster paper speed equals more resolution. Chart recorders were around long before swept filters which were among the first tools for frequency analysis.

 

[SomptingGuy]

You need to be equally happy in both domains. In the domain of rotational machinery, I'd always start in the frequency domain. But I'd insist on a Campbell diagram (or colormap, waterfall, etc) to begin with. Then you get to see the whole behaviour in a single picture. Local behaviours can be found and studied in more detail by slicing through the colormap.

Specifics can be found by viewing small sections of data in the time domain.

Many P&E engineers are spooked by anything to do with frequency domain analysis, so you have to keep reminding them that a spectrum is just another way of describing the shape of a time (or crank angle) domain signal.

- Steve

 

[MikeyP]

At the risk of flogging a dead horse...

That's all well and good, but I think the OP is interested more in how you simulate time domain signals rather than display them.

I would also infer that the OP is interested in looking at pressures inside the system rather than radiated noise.

M

--
Dr Michael F Platten

 

[SomptingGuy]

Hmm, I agree. I usually associate the word "analysis" with processing of measure (or simulated) data. If by "analysis" he means "simulation", that's a different matter entirely.

If one is attempting to simulate the behaviour of pressures inside the system, or noise radiated from orifices due to velocity fluctuations, then non-linear, time-domain simulation has been standard practice in the automotive engine industry for c. 15-20 years. Some die-hards still swear by their linear, frequency-domain models, but the tide has long turned.

Non-linear time-domain models can (obviously) capture non-linear effects (high SPLs, mean flow, temperature gradients). They can also simulataneously predict acoustics AND performance. It's easy to reduce radiated noise in isolation, but how will your measures afffect back pressure?

- Steve

 

[hacksaw]

time domain offers little insight especially when it comes to characterizing system modes or linking system specific design changes with response changes.

frequency domain methods are a very good tool if properly used.

 

[DLite30]

Just to clarify my original post, I meant to indicate how these apply to design simulation of acoustic systems...not field analysis.

I believe we will continue to allow the use of both methods of simulation...linear, frequency domain, and non-linear time domain.

As we have installed many recip machines from 1000 to 8000 hp using frequency domain-based designs which have worked well so far.