Railway noise modeling 3

[juanuc]

Hi everybody in the forum:

I've just began a project concerning noise control in trains. The aim of the project is develop a model predicting vibration and noise of the system, trying to eliminate it from the design stage. As far as I'm concerned, I'm not sure if it's a good idea trying to model such a complex problem, or if it should be more realistic using some comercial or "self-made" software. What do you think about that? Do you have any expertise that could be useful for me? Any help?


Thank you very much for your tips.

 

[nbucska]

Are you concerned about the noise on board or in the vicinity of the rail?

In either case I think modeling is nearly impossible. You can model only systems what you are able to adequatelly describe.

If you want to reduce the noise I think your best chance is to do it experimentally.



<nbucska@pcperipherals.com>

 

[juanuc]

I'm concerned about noise on board and in the vicinity of the rail too, and the point is to predict it with no experiments. In fact we have no train, is a new development, and that the reason why we planned to get a model is to avoid the need of expensive experiments.
We are going to work directly with design engineers, helping them with a new point of view, so we do know the description of the system.

 

[butelja]

I have no experience with this software package, but there is a software called AUTOSEA2, which uses statistical energy analysis to predict interior noise levels. The brochure describing the software is available at

http://www.vasci.com/products/autosea2/autosea21.pdf

.

 

[GregLocock]

SEA is a fine technique... but I doubt you'd get within 20 dB of the real answer if the original poster really wants to persevere without doing any experiments.

SEA also only works where the modal density is reasonably high (say > 3 mode/octave, and really you need a lot more than that).

I work in the auto industry. FEA based acoustic prediction works up to 100 Hz, reliably, if you can get the exciation forces, which we typically derive experimentally. We push this to 200 Hz and cross our fingers.

On a car SEA works quite well down to 500 Hz. Note that SEA needs a lot of experimental data to back it up.

The gap between 100 and 500 Hz is of course where most of our A weighted noise problems come from!

Cheers

Greg Locock


Cheers

Greg Locock

 

Hello,
If you are interested in noise outside a train, I can tell you there is a lot of information about. Almost every western country has developed it's own methode of calculating noise next to the track. This offcourse radiates noise into the train. Maybe this is useful in your model

 

[fab]

Hallo,
I'm working on a european project whit the aim to predict the pass-by noise (at 20 m from the track) for high-speed trains.
The software named &quot;TWINS&quot;, developped by Dr. Thompson (ISVR, Southampton) seems to work well, with a good agreement with experimental measurements...

Fabrizio

 

[KOKOKOKO]

I wonder why in modelling the Rail-way vibration, the model consists of one concentrated harmonic Load.

 

I`d read some orientations about vibrations modeling in FTA`s manual, but it is recommended measure ome characteristics of the propagation and acting forces, (trin forces in this cases). The vibrations modeling is too complex and depend strongly of the soil properties and structure, (subterrain channel waters can be a great problem).
In the noise case, the modeling is too much simply; there are a lot of softwares and procedures very clear. FHWA and FTA gives gudelines in this topic.

 

[MikeyP]

I'm sure I have seen papers which have successfully modelled rail carriages with SEA. If you have access to scientific literature databases then try including &quot;SEA&quot; or &quot;Statistical enegy analysis&quot; in your search terms. SEA can work well as long as you conform to its assumptions. The software can be very expensive, but an accurate solution probably requires a sophisticated model which includes double walls. These problems are not usually suitable for homegrown software.

BTW, Commercial SEA software often lags someway behind what is out there in the scientific literature. There are some pretty clever models available if you know where to look and how to implement them. Greg's mid-frequency problems are slowly being addressed too. Try a search on &quot;SEA&quot; and &quot;Resound&quot;.

Michael

 

[nametab]

I've worked on railway noise within the railway industry for 20+ years, so with the danger of sounding like an old .... Also, the original question is some months ago, so perhaps the original topic is closed now.

External noise at speed, being almost exclusively dependent on wheel-rail interaction, is very well modelled by TWINS, as already discussed.
I concur with the previous comments about SEA. An excellent tool IF IF you know what you're doing, and understand the realities of its modelling constraints.
For internal noise in general, anyone thinking they can model from scratch, without experiment or existing supporting data, is deluded. The topic and behaviour is extremely complex, and dependent on many parameters and variables. Structure-borne contributions in particular are poorly quantified and modelled. Hundreds of man years have been spent so far across the railway industry and by consultants in developing predictive methods, and really there is comparatively little to show for it. A very few specialist consultants have got reasonable predictive skills now for railway vehicles, but they are inevitably a mix of model, data, empirical adjustments and experienced application.