 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Natural Science Forum / Physics / Acoustics / August 2005Barrier Attenuation Modeling and Noise BarrierHeight Design
Dear Acoustikers: For some decades, I have been using a barrier attenuation model to predict the height required for effectively screening housing sites from highway noise. Land developers have usually gone along with the required heights as long as they are economically feasible, but they also "push the envelope", and I try to accommodate they economic wishes to the extent we will fine-tune barrier heights and residence grades to optimally scoop out just enough soil to form the barrier. And in that mode, every foot matters. Anyway, a recent project has particularly pushed my envelope in that at 20 feet barrier height, "justification" will be relatively important. So I revisited the old FHWA barrier attenuation model, as well as some JASA articles on barrier performance in general. Three aspects emerge as significant: 1- They all have supporting data for Fresnel numbers in excess of 1.0. 2- There is a decided absence of data on small barrier heights, i.e. where the barrier summit is just above or just below the line of sight. This is the most common situation that we (well, maybe just me) face day to day. Here it is most important that intelligent decisions are made. 3- There is a difference between the attenuation expected for a line source (traffic) and a point source (HVAC fan). In re 2-, the classical (Maekawa) attenuation for zero line-of-sight (LOS) break is a 5 dB value. Yet intuition says it ought to be 3 dB. The common value for a point source is something else. In surveying reports on the fit of field test data to various prediction models, it is often the case that Pierce's formulation (Figure 9-27), which concentrates on interference of the two waves off the edge interfering with one-another seems to get better agreement. Furthermore, for barriers of small height, the ground absorption is very important. It is problematic to model, since theoretical treatments may want to deal with sound power and surface absorption, while practical design problems start with existing SPL at a receiver site, and want to design to a new status with a barrier in place and a final ground absorption that cannot be changed from a fixed residential choice. So what I am looking for are: Zero LOS modeling attenuation value for line and point sources separately (is it 3, 5 or 6 dB?). Attenuation vs Fresnel number, line and point source separately, that correctly meld to the zero LOS benchmark. Negative Fresnel number range (ground under LOS). Any suggestions? Thanks, Sincerely, Angelo Campanella. --------- www.CampanellaAcoustics.com --------- [snip] > In re 2-, the classical (Maekawa) attenuation for zero line-of-sight >(LOS) break is a 5 dB value. Yet intuition says it ought to be 3 dB. The >common value for a point source is something else. [snip] I presume your intuition is based on thinking of ray acoustics and sharply delimited shadow zones. Since propagation over a barrier implicitly involves diffraction, that kind of geometric intuition is misleading. In real life, a zero LOS barrier geometry is rarely zero. Wind and temperature gradients move it around from second to second, so it's really hard to count on what you'll get. If you've lost ground attenuation, then the barrier can make things worse. > Furthermore, for barriers of small height, the ground absorption is >very important. It is problematic to model, since theoretical treatments >may want to deal with sound power and surface absorption, while >practical design problems start with existing SPL at a receiver site, >and want to design to a new status with a barrier in place and a final >ground absorption that cannot be changed from a fixed residential choice. If you want to refine things in barrier situations where ground effect still plays a role, look at the GTD analysis that Rasmussen has put together. Merger of the Pierce wedge type analysis with the reflected paths from the ground. The old FHWA procedure, based on Maekawa's curve and some simple rules, really loses a lot of detail in complex situations. Ken Plotkin > I presume your intuition is based on thinking of ray acoustics and > sharply delimited shadow zones. Since propagation over a barrier > implicitly involves diffraction, that kind of geometric intuition is > misleading. Yes. That's why I sent the message! > In real life, a zero LOS barrier geometry is rarely zero. Wind and > temperature gradients move it around from second to second, so it's > really hard to count on what you'll get. If you've lost ground > attenuation, then the barrier can make things worse. The geometric line of sight break can be and is frequently "near zero". Call it a crude ground effect problem if you wish, but it is a very common situation in real world residential property design as we develop vacant land, and the project is highly budget driven. In particular, to evaluate and design proposed barriers, it is most practical to field-measure the existing highway or other noise source DNL value, then it becomes necessary to model the "before" scenario with the chosen propagation software "model", then to model proposed barrier solutions with the same software model. Most of the time, the existing terrain is exposed ("negative" or "near zero" LOS break). Call it foolishness, but there is no law against fools (just bad luck for the unfortunate among them!). > If you want to refine things in barrier situations where ground effect > still plays a role, look at the GTD analysis that Rasmussen has put > together. Might you furnish a reference? > Merger of the Pierce wedge type analysis with the reflected > paths from the ground. The old FHWA procedure, based on Maekawa's > curve and some simple rules, really loses a lot of detail in complex > situations. I agree with the details loss, Apparently, there is an alternative list of parameters that must be furnished (other than observer & source heights and LOS break). Those I need to know. The proper way to proceed, and that which I pursue, is to measure the existing DNL (easy to do these days), model the existing terrain including a DNL match, then feed the same model with variable values that represent practical barriers, then massage them to get DNL65 max for ten years hence, and then campaign to get that solution fleshed out by the developer. I have done all of that except the refinement around "near zero" LOS break I seek. I am now downloading the documents that Kari suggested. SP05 is mainly about sources. VT104 is about noise emission from a myriad of vehicle types. SP10 is about source position and height modeling. All seem to presume that the barrier attenuation models used in various countries need not be questioned. Angelo Campanella, > The geometric line of sight break can be and is frequently "near zero". Yes, when you draw the geometry as straight lines it often is. >Call it a crude ground effect problem if you wish, but it is a very >common situation in real world residential property design as we develop [quoted text clipped - 7 lines] >there is no law against fools (just bad luck for the unfortunate among >them!). [snip] Don't forget the infamous Toronto barrier error in the early 70s. As far as budget driven design goes, I've never had trouble designing barriers for townhouse developments that were way cheaper than what a highway department would design...big difference being that the barrier can be on the developer's property, and can usually wrap around. But I've never submitted a design that was zero LOS. Even if the numbers come out right, the eventual homeowner is never going to believe that he's getting any benefit if he can still see the source. > Might you furnish a reference? Rasmussen, K.B., "The Effect of Terrain profile on Sound Propagation Outdoors," Danish Acoustical Institute Report 111, January 1984. Rasmussen, K.B., "Outdoor Sound Propagation Near Ground Surfaces," The Acoustics Laboratory, Technical University of Denmark Report no. 45, 1990. The second is his thesis, and includes copies of journal publications. Not in the office today, but if you do a lit search on Karsten Bo Rasmussen you should turn those up. BTW, that stuff is really solid. He conducted careful measurerments as he was developing the theory. Those were over distances of interest for things like highway noise. The algorithms work well over longer distances and rugged terrain (like the Grand Canyon see http://overflights.faa.gov/apps/GetFile.CFM?File_ID=22 } and are incorporated in NMSim ( www.nmsim.com - should be downloadable shortly, once our webmaster wakes up and puts the distrubution on the site). With the addition of refractive effects, this is also the core of Nord2000. There were also publications by Jonassen (think I've spelled it right) in JASA in the 80s (more or less) on the combination of barriers and ground effect. > The proper way to proceed, and that which I pursue, is to measure the >existing DNL (easy to do these days), model the existing terrain [quoted text clipped - 3 lines] >by the developer. I have done all of that except the refinement around >"near zero" LOS break I seek. How will you model the existing terrain? For Maekawa thin screen shielding at zero or negative LOS break, don't forget that the formula kicking around since the 70s uses an analytic fit that's good at positive LOS break but pretty much sucks at low and negative values. I've always used Maekawa's orginal curve. > I am now downloading the documents that Kari suggested. > > SP05 is mainly about sources. > VT104 is about noise emission from a myriad of vehicle types. > SP10 is about source position and height modeling. I just get "This page cannot be found" for those. Do you think you sapped Kari's bandwidth allocation? :-) > All seem to presume that the barrier attenuation models used in various >countries need not be questioned. Typical EU approach to things. Ken Plotkin ...clip... > There were also publications by Jonassen (think I've spelled it right) > in JASA in the 80s (more or less) on the combination of barriers and > ground effect You may refer to Hans Jonasson's paper (based on his PhD thesis) 1) Sound reduction by barriers on the ground, J. Sound Vibr. 22(1972) 113-126. Thesis: The propagation of sound over ground with and without barriers report 18, Lund University, Sweden, 1971. This prediction method is different from those ordinary ones. The method make use a secondary sound source at the upper edge of a barrier. A good summary of "the state of the art" in 1970's is Kurtze U J, Noise reduction by barriers, J. Acoust. Soc. Am. 55(1974)3, 504 -518. >> The proper way to proceed, and that which I pursue, is to measure the >>existing DNL (easy to do these days), model the existing terrain [quoted text clipped - 12 lines] > I just get "This page cannot be found" for those. Do you think you > sapped Kari's bandwidth allocation? :-) Sorry. The www-site seems to bee down. Another site/page that points to these papers is http://www.imagine-project.org/artikel.php?ac=direct&id=289# But these links also show that http://www.harmonoise.nl is down. If you are interested in these papers I can email these to you. Kari Pesonen >You may refer to Hans Jonasson's paper (based on his PhD thesis) >1) Sound reduction by barriers on the ground, J. Sound Vibr. 22(1972) > 113-126. >Thesis: The propagation of sound over ground with and without barriers > report 18, Lund University, Sweden, 1971. Yes! That's it. Sorry about not spelling his name right...just going from memory. >This prediction method is different from those ordinary ones. The method >make use a secondary sound source at the upper edge of a barrier. Yes. Rasmussen's work follows in those footsteps. Very powerful method. >Sorry. The www-site seems to bee down. Another site/page that points to >these papers is >http://www.imagine-project.org/artikel.php?ac=direct&id=289# >But these links also show that http://www.harmonoise.nl is down. > >If you are interested in these papers I can email these to you. Thanks for the offer. But I've saved the links, and I'll try again in a few days. Sites eventually come back up. Ken Plotkin > Don't forget the infamous Toronto barrier error in the early 70s. Refresh our memories on the brief details. >> Might you furnish a reference? > Rasmussen, K.B., "The Effect of Terrain profile on Sound Propagation [quoted text clipped - 3 lines] > Acoustics Laboratory, Technical University of Denmark Report no. 45, > 1990. > Not in the office today, but if you do a lit search on Karsten Bo > Rasmussen you should turn those up. BTW, that stuff is really solid. Thanks > There were also publications by Jonassen (think I've spelled it right) > in JASA in the 80s (more or less) on the combination of barriers and > ground effect. It's Hans Jonasson. I know Hans well; we're collaborating on sound power standards these days. He did not get to Fresnel numbers less than 1. This hiatus is common in all early experimental studies. Just seems to indicate the confusion that exists below 1.0 Pierce is among the first to tackle it. And he has to lean on coherent analysis (interference effects) to get anywhere there. The Fresnel number value range needs to include .01 to 1.0 to make useful barrier cost estimates and tradeoffs for "before and after" predictions for a homespun barrier. > How will you model the existing terrain? Somewhat nil so far except a simplistic terrain absorption value by the source and by the receiver. A barrier is then added in the model. Needs further work (this thread). > For Maekawa thin screen shielding at zero or negative LOS break, don't > forget that the formula kicking around since the 70s uses an analytic > fit that's good at positive LOS break but pretty much sucks at low and > negative values. I've always used Maekawa's orginal curve. That says 5 dB at zero LOS break! And how does one distinguish line sources from point sources? > I just get "This page cannot be found" for those. Do you think you > sapped Kari's bandwidth allocation? :-) I have them on HD if you need them. > Typical EU approach to things. We'll work on it! Ang. Campanella >> Don't forget the infamous Toronto barrier error in the early 70s. > >Refresh our memories on the brief details. They made roadside measurements around some highways. Designed barriers by using Maekawa's formula, and subtracting that from the measured levels. The barriers underperformed. That's when everyone realized that when you put in a barrier you usually lose most (if not all) of the ground attenuation. >It's Hans Jonasson. I know Hans well; we're collaborating on sound power >standards these days. Then you should be picking his brain. >He did not get to Fresnel numbers less than 1. This hiatus is common in >all early experimental studies. Just seems to indicate the confusion [quoted text clipped - 3 lines] >make useful barrier cost estimates and tradeoffs for "before and after" >predictions for a homespun barrier. There is nothing wrong with Maekawa's original curve in that range. Just with the widely used analytic fit. >> How will you model the existing terrain? > >Somewhat nil so far except a simplistic terrain absorption value by the >source and by the receiver. A barrier is then added in the model. Needs >further work (this thread). As I've said, you can't just add barrier shielding by itself to the no-barrier ground attenuation. If all you're doing for the barrier is Maekawa, the old FHWA procedure of taking the larger of the two is really appropriate. >That says 5 dB at zero LOS break! And how does one distinguish line >sources from point sources? Why do you doubt the 5 dB? That's what diffraction theory says, and that's what's measured in the laboratory. As far as point from line..,while we think of highway noise as a line source, it's a distribution of uncorrelated point sources. Very different from a true line source like an exploding wire. But easier to deal with. Kurze and Anderson had a paper in the early days (1970ish) in (if memory serves) JSV. Just an integration of a distribution of uncorrelated point sources, using Maekawa for each source position. That is the basis or the line model in the old FHWA noise model, and partial sums from that model are the partial barrier tables. Ken Plotkin > They made roadside measurements around some highways. Designed > barriers by using Maekawa's formula, and subtracting that from the > measured levels. The barriers underperformed. That's when everyone > realized that when you put in a barrier you usually lose most (if not > all) of the ground attenuation. Been there. done that. More likely that the ground attenuation has to be figured in for modeling, is is what I've been doing. > Then you should be picking his brain. Your information has just recently advised me of that connection. I have not yet broached the subject. > There is nothing wrong with Maekawa's original curve in that range. > Just with the widely used analytic fit. I have seen some others' rendition of his work, and his attenuation vs Fresnel number(?) by their rendering goes though Fresnel=zero and so forth, I have a hard time relating that to site variables. > As I've said, you can't just add barrier shielding by itself to the > no-barrier ground attenuation. If all you're doing for the barrier is > Maekawa, the old FHWA procedure of taking the larger of the two is > really appropriate. The larger of the two which? > Why do you doubt the 5 dB? That's what diffraction theory says, and > that's what's measured in the laboratory. Fair enough at this point. Allan Pierce's brief treatment is for a point source, and his methodology of treating the ground reflection leads straight to interference effects, which do not show in Maekawa's method. This makes for confusion. which is correct? > As far as point from line..,while we think of highway noise as a line > source, it's a distribution of uncorrelated point sources. Very > different from a true line source like an exploding wire. But easier > to deal with. For uncorrelated noise sources, some interference effects will wash out, so that may be why FHWA still lives on with Maekawa. > Kurze and Anderson had a paper in the early days > (1970ish) in (if memory serves) JSV. Just an integration of a > distribution of uncorrelated point sources, using Maekawa for each > source position. This is the offset which I think is attempted to be represented in Cyril Harris's (1979 edition) text Figure 3.5. (This Chapter 3 is by J.E. Piercy and Tony Embleton.) The Fresnel #o.01 (vanishing) value shown there gives 5 dB for an elevated point source. 4 dB for an elevated and incoherent line source, and zero dB for a point source on the ground. Now do you understand my quandary? I think I might should be using a vanishing value of 4 dB instead of 5.... or go to a modification of Allan Pierce's point source method... or study the whole thing from scratch... not likely soon at this pace... Cheers, Ang. C. > Fair enough at this point. Allan Pierce's brief treatment is for a >point source, and his methodology of treating the ground reflection >leads straight to interference effects, which do not show in Maekawa's >method. This makes for confusion. which is correct? Maekawa's work is for a barrier in a free field. Since Pierce's analysis includes the ground that is a different problem. Both are correct. Neither corresponds to the particular geometry you have. > For uncorrelated noise sources, some interference effects will wash >out, so that may be why FHWA still lives on with Maekawa. We're dealing with hourly averages. "Washing out" is a rather negative way of portraying that. The basic integration of a moving point to get a line source - which is the basis of all Leq highway models - is based on that source not being time autocorrelated with itself. > Now do you understand my quandary? I think I might should be using a >vanishing value of 4 dB instead of 5.... or go to a modification of >Allan Pierce's point source method... or study the whole thing from >scratch... not likely soon at this pace... Yes. You just aren't up to speed on the genral treatment of a barrier in the presence of the ground. You should really look up Jonasson's papers. Ken Plotkin Wind Turbine Noise 2005 First International Conference on Wind Turbine Noise: Perspectives for Control Will be held in Berlin, Germany on 17th and 18th October 2005 Registration now open. See www.windturbinenoise2005.org Organised by INCE/Europe ____________________________________ > First International Conference on > Wind Turbine Noise: Perspectives for Control Interesting topic. Any suggestions for web locations of field data on wind turbine noise, especially full spectra (down to 10 Hz), preferably in 1/2- octave bands, at distances of 100', 1,000', and 10,000 ft (30m, 300m, 3000m)? I know that "it varies widely with size and type". Just state the machine type. Angelo Campanella >> First International Conference on >> Wind Turbine Noise: Perspectives for Control [quoted text clipped - 8 lines] > > Angelo Campanella Here are some American documents, but not exactly what you are looking for http://www.nrel.gov/docs/fy04osti/34478.pdf http://www.nrel.gov/wind/pdfs/34021.pdf http://www.nrel.gov/wind_meetings/fy05_meeting/pdfs/migliore_fy2005.pdf Kari Pesonen Is the noise these things generate infrasonic? Yikes, elephants in India will be trying to talk to it while all the locals get sick. <Dr. Vladimir Gavreau> John > Wind Turbine Noise 2005 > First International Conference [quoted text clipped - 8 lines] > Organised by INCE/Europe > ____________________________________ > Dear Acoustikers: > > For some decades, I have been using a barrier attenuation model to predict > the height required for effectively screening housing sites from highway > noise. ... ...clip... I point only one issue: Many highway noise/barrier prediction models/methods expect only one or two source heights, although sound radiation from traffic is vertically and horisontally distributed, and the distribution depends, for example, on speed and road pavement. These Harmonosie papers give some more information about ideas: http://www.harmonoise.nl/bestanden/D07_WP1.1_HAR11TR-020614-SP05.pdf http://www.harmonoise.nl/bestanden/D08_WP1.1_HAR11TR-030108-VTI04.pdf http://www.harmonoise.nl/bestanden/D09_WP1.1_HAR11TR-041210-SP10.pdf Kari Pesonen Angelo Thanks for asking a question that has probably given me enough reading material of the next 6 months. Although I don't have the technical nouse to give a good, I am all too fimilar with the application of pressure by developers. Looking forward to any sight testing you can do to confirm the method you eventually use. Tom I have recently returned to the field of acoustics after an absence of 25 years and read this thread with some interest and slight bafflement which may well be ignorance about practicalities on my part. Is there some reason a straightforward numerical simulation will not provide the information you seek? > I have recently returned to the field of acoustics after an absence of > 25 years and read this thread with some interest and slight bafflement > which may well be ignorance about practicalities on my part. Is there > some reason a straightforward numerical simulation will not provide the > information you seek? Yes: Mathematics (in physics) is not reality. It is a but a simulation of reality. Only when it has been demonstrated by experiment of real world field measurements can there be confidence that the mathematical simulation really and properly represents that which is inferred, can we then accept that mathematical model is valid. This is true in any technology. Acoustical phenomena, often not in themselves being visible, is particularly difficult to be quantitatively evaluated that way. Real world problems... ugh! Angelo Campanella I seem to have phrased my question poorly. I cannot access the Harmonoise site either and nobody here seems to have the final report so I am flying a bit blind. I agree fully that an engineer needs to understand the strengths and limitations of the tools they are using but these are significantly different for numerical simulations (e.g. BEM and FEM) and classical 'lumped' analytical and semi-analytical approaches. I had assumed you were using the latter (your response is consistent with the use of the latter) and my question was about the use of the former. Please correct me if I am wrong but I believe you wish to consider a situation where the source is known and the problem is to solve the propagation of acoustic waves. The accurate numerical simulation of acoustic sources is a tough (or a very large) problem but the propagation of acoustic waves can be predicted accurately for many engineering purposes on a modern PC. A barrier and a few buildings would appear straightforward if the source is specified. Perhaps I am missing something obvious given that I have not solved such a problem myself. Do engineers have wide access to such software? > I seem to have phrased my question poorly. I cannot access the > Harmonoise site either and nobody here seems to have the final report > so I am flying a bit blind. I can e-mail a copy if you wish. > 'lumped' analytical and semi-analytical approaches. I had assumed you > were using the latter (your response is consistent with the use of the > latter) and my question was about the use of the former. My question relates to the validation data; I have not seen good validation data for small barrier heights. I want to be able to use monitoring data I have already taken on-site to accurately designate the barrier height required to reduce the sound to a designated level. In my case, I need to reduce DNL 77.9 traffic noise to be not more than DNL 65 for a 5' tall observer within, say 50' (15m), of that sound barrier. I do not expect to have to designate a greater barrier height "just to be safe", as that costs my client a lot of money, and may squelch the project. As you may know, the cost of a barrier wall is likely proportional to the square or the cube of its height. Figuring 100mph+ winds that may occur in Ohio, you can see that this is not a trivial matter. When the project is built, I expect that indeed DNL 65 noise is not exceeded by traffic on the average for at least ten years, that this should be the case with sufficient confidence that I can sleep well at night. On top of that, we will be be tapering the wall on a line perpendicular to the traffic as the development extends away from the traffic, on either side of the development, and that extension will ultimately involve a lower and lower barrier wall. > Please correct me if I am wrong but I believe you wish to consider a > situation where the source is known and the problem is to solve the > propagation of acoustic waves. The accurate numerical simulation of > acoustic sources is a tough (or a very large) problem but the > propagation of acoustic waves can be predicted accurately for many > engineering purposes on a modern PC. True, but the PC must be furnished with a valid and validated mathematical model. What has raised my suspicion is that the commonly stated 5 dB attenuation for a zero line-of-sight break has mathematical rigidity, and likely model laboratory measurements to verify it, but those experimental verifications have been performed with a point source of sound (e.g. single loudspeaker). There is an inference that for a line source (e.g. a continuous stream of vehicle traffic parallel to the barrier wall), that the effective sound reduction of that barrier wall for a zero line-of-sight elevation above the barrier wall is less than 5 decibels. So far, I have not seen proof or disproof of that contention. It may have been done and published, but I have yet to see it with my own eyes. Seeing is believing. Since I do not have a large library at my fingertips, I cannot verify that today, or perhaps not even tomorrow. When I went through my JASA 1929-1996 CD set (the years over which validation was reputed to have been performed), I found several experimental studies, but they all performed where the line-of-sight break was quite positive, e.g. good sound barriers. None worked with the marginal situations (that we face quite commonly around residential construction sites these days). Virtually all modeling simply assumes that the 5 dB point source value is universally true no matter what the sound source is. I may be wrong in this interpretation, but I really like to, in the case of engineering application for public use, reverify these "truths" from time to time. Angelo Campanella > My question relates to the validation data; I have not seen good >validation data for small barrier heights. I want to be able to use You won't. Performance is way too variable. >monitoring data I have already taken on-site to accurately designate the >barrier height required to reduce the sound to a designated level. In my >case, I need to reduce DNL 77.9 traffic noise to be not more than DNL 65 >for a 5' tall observer within, say 50' (15m), of that sound barrier. I >do not expect to have to designate a greater barrier height "just to be >safe", as that costs my client a lot of money, and may squelch the project. If you had a thousand pound load to lift, what strength cable would you use? If you had two such loads, one on solid land, and one on a ship in rough seas, would you choose the same cable? Is "just to be safe" different than "to ensure that the requirement is met"? [snip] > Virtually all modeling simply assumes that the 5 dB point source value >is universally true no matter what the sound source is. I may be wrong >in this interpretation, but I really like to, in the case of engineering >application for public use, reverify these "truths" from time to time. That's not an assumption. That is what happens with diffraction of free field sound by a thin screen barrier. It's not true that "virtually all modeling" does that. Maybe you've been looking at work done by people who grab a formula without understanding what it's good for. If you're trying to shave the design as far down as possible, you need to escalate the level of analysis. Try getting a copy of Nord2000, which accounts for barriers, terrain and atmospheric refraction. Get wind data for your site - you can download hourly surface data and twice-daily upper air profiles for free from NOAA's NCDC site. Run your analysis hourly for a couple of year period to get decent statistics, and size the low barrier based on that. Ken Plotkin > I can e-mail a copy if you wish. Thanks that would be appreciated but I have a couple more people to try who ought to have a copy. If I fail I will drop you an email. > My question relates to the validation data; I have not seen good > validation data for small barrier heights. As mentioned earlier, this has not been my area but others in the department have posters up with measurements and predictions of barriers. The thrust of much of the research appears to be the development of devices ontop of the barrier to allow a lower height. There are anechoic and field measurements plus some analytical/semi-analytical predictions. The latter are 2D and likely to be a bit limited for engineering purposes but it was not a large component of the work. It would be a straightforward task to repeat using a general numerical method if it was of interest to the industry. For a loudspeaker, barrier and microphone in still air the resulting predictions are almost certain to lie within experimental tolerances. I would have assumed this had be done by many researchers before. Is this not the case? For larger cases involving buildings, uncertain temperature and wind velocities, imprecise impedances such as the ground, trees and weird devices on top of the barrier there will be a greater discrepancy between measurements and the predictions because of the guesses the user would have to make about the unknown input parameters. Had the user been able to specify all input parameters exactly and used a sufficiently fine computational grid the accuracy of the predictions would be determined by the governing equations which are likely to be comparable and probably exceed experimental tolerances for most engineering cases. This is quite different to the analytical/semi-analytical methods because the user is largely in control of the achieved accuracy. > True, but the PC must be furnished with a valid and validated > mathematical model. For lumped models based on experiments I would agree. For numerical methods which accurately simulate the physics then, in my experience, it is mainly the user which needs validating. The performance of the numerical scheme can be fully defined by the equations it solves, the numerical scheme it uses and the boundary conditions it supports. > Virtually all modeling simply assumes that the 5 dB point source value > is universally true no matter what the sound source is. This has no meaning for numerical schemes which simulate the physics without calling on such assumptions. > The thrust of much of the research appears to be the > development of devices ontop of the barrier to allow a lower height. Papers I have attended seemed to show that absorption placed there, especially on a horizontal surface gives a dB or so of improvement. One Japanese author proposed a cylindrical topping for a barrier, with perforations to make a weatherproof improvement. It can be added to any barrier with minimal consequences. > There are anechoic and field measurements plus some > analytical/semi-analytical predictions. The latter are 2D and likely to [quoted text clipped - 5 lines] > would have assumed this had be done by many researchers before. Is this > not the case? At Ken Plotkin's persistent prodding, I printed out and am now reading Kurtze's 1974 JASA paper. There, in discussions of Figure 4 (Makawa's chart), and on page 508 in discussing interference effects, which may be significant at lower frequencies and small barrier heights, Kurtze concludes that the Maekawa curve overestimates barrier attenuation by 0dB for N>>1, and up to -3 dB for N=0. This is the kind of shortfall I have been alluding to. It argues for the concept that, given a choice of no barrier or a low barrier (in the vicinity of the line-of-sight height) it may be zero-sum game since interference can make the noise at the receiver location greater, especially for some lower frequency pure tones. > For larger cases involving buildings, uncertain temperature and wind > velocities, imprecise impedances such as the ground, trees and weird [quoted text clipped - 6 lines] > comparable and probably exceed experimental tolerances for most > engineering cases. All of this points to the need for a philosophy for the "Minimal Effective Barrier". For chickens, I know that to keep then corralled, the fence has to be just higher than they can jump while flapping their wings... about 3 feet as an absolute minimum; 4 feet is advisable. For roadway noise barriers, there needs to be a similar criterion. This relates especially to designing noise barriers around housing developments where the outdoor yards need to be protected to be below DNL65. Yet to build lengths of barrier, the cost is still a factor, not mention aesthetics. Thus, with a 3 dB of uncertainty on account of interference, one conclusion is that it is fair to continue a barrier into the DNL64 region for instance, at a height that is above the line of sight to traffic, but then to terminate it abruptly where the existing ambient has reduced to, say, DNL63. Well... this is a start, anyway..... Angelo campanella [snip] >propagation of acoustic waves. The accurate numerical simulation of >acoustic sources is a tough (or a very large) problem but the >propagation of acoustic waves can be predicted accurately for many >engineering purposes on a modern PC. A barrier and a few buildings >would appear straightforward if the source is specified. [snip] Prediction of sources is actually not that difficult. In this kind of problem, source definition is emprical. Measurements are done under relatively simple, controlled, conditions. Enough repetitions are performed to get statistically reliable results. Propagation is the hard part. The atmosphere is not homogeneous, and varies from second to second. Propagation from a source within a few feet of the ground to a receiver a few feet from the ground is very sensitive to changes in refraction. Propagation over a low barrier (as Ang wants to do) is very nasty. You know how the knife edge in a Schleiren system greatly amplifies things relative to a shadowgraph? Well, a low barrier where the ground still plays a significant role has great similarities. Ken Plotkin > Prediction of sources is actually not that difficult. We cannot be talking about the same thing. The accurate simulation of, say, squeal from a single break disc is a large computation and one would still need to call on a degree of statistical modelling for the upper frequencies. If you add the noise from engine and powertrain and the near field modelling around the vehicle the computational requirements become prohibitive. Yes there are enormously simpler models for the sources but then we are back to the problems of lumped parameters and large degrees of empiricism. I should add that I am not criticising the approach because it is what is required to actually perform simulations today. > Propagation is the hard part. It is not trivial but would appear viable on a modern PC. I would classify propagation in still, isothermal air as straightforward. Propagation in the presence of constant wind and temperature variations as a bit more difficult to implement and slower to run but still accurate. The accuracy will take a hit if the effects of turbulence are modelled statistically or else the computational cost will jump greatly if the turbulence is solved numerically. The latter can be performed on a massively parallel computer over a period of days but I would argue that it would only need doing if something had been observed that could not be explained/understood. Since I have not performed these simulations I cannot make the above statements with absolute authority but I cannot see what relevant physics would be missing from the simulations. But I am open to suggestions? >We cannot be talking about the same thing. The accurate simulation of, >say, squeal from a single break disc is a large computation and one >would still need to call on a degree of statistical modelling for the >upper frequencies. If you add the noise from engine and powertrain and >the near field modelling around the vehicle the computational >requirements become prohibitive. Definitely not talking about the same thing. Brake squeal is very complex, and not particularly repeatable. The subject here is highway noise. The noise sources are entire vehicles, we are looking to predict an hourly average, and there are enough of them to get a meaningful statistical average. Sources are determined empirically. There is an analogy with jet aircraft. If you poke around the NASA centers, you find models like ANOPP that predict engine noise from basic physics. Very complex, very highly developed. But in the models that are used to predict noise around airports, the sources are empirical, based on setting microphones out while the aircraft does controlled flyovers at various power settings and configurations. When it comes down to computing contours around an airport, there is considerable uncertainty because the modeler (i.e., the person entering data into the noise model) generally has poor information as to where the aircraft are and what the pilots are actually doing. >It is not trivial but would appear viable on a modern PC. I would >classify propagation in still, isothermal air as straightforward. [quoted text clipped - 6 lines] >that it would only need doing if something had been observed that could >not be explained/understood. Still and isothermal is trivial. Propagation through wind and temperature gradients isn't really that hard. I've written software that does that, and it works just fine. As far as running time, on a PC I've computed the hourly noise for a full year of hours at an airport in about two weeks of crunching on a single 2.something GHz machine. That accounted for wind and temperature gradients, plus a bit of turbulence and diffraction,. The problem is that you don't know what the gradients will be at any particular time in the future. And unless you've done the kind of long-term retrospective crunching like I mentioned above, you don't really know what the statistics will be. >Since I have not performed these simulations I cannot make the above >statements with absolute authority but I cannot see what relevant >physics would be missing from the simulations. But I am open to >suggestions? It's not the physics. It's the foresight. Ken Plotkin |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.