Problems with TNM 3.0

Transcription

1 Problems with TNM 3.0 from the viewpoint of SoundPLAN International LLC TNM 2.5 TNM 2.5 had some restrictions that hopefully are lifted in the up-coming version of TNM 3.0. TNM 2.5 for example did not allow an elevation line to intersect with a road or a noise barrier. In the past, models were small and it was relatively easy to clip elevation lines at the edge of the road, with the vast amount of data typically used in GIS operations, this may be a lot more cumbersome. Other problems such as the 2D spreading for sure are resolved but some of the problems remain. The assessment of TNM errors is based on the TNM calculation kernel from July It is our understanding that the Consistency Test Suite is using the same version to produce the master results. In May 2015 SoundPLAN sent the Volpe Center our first feedback on the results of the Consistency Test Suite. SoundPLAN has no information if these evaluations so far led to changes in the TNM coding. Problem #1 : Geometrical Spreading TNM up to its version TNM 2.5 calculated the spreading only with 2 dimensional coordinates, thus only taking the horizontal distance into account. Receivers that were close to a road but high above it experienced basically the same noise levels as a receiver close to the ground. TNM 3.0 has introduced 3 dimensional coordinates for the calculation of the spreading but now calculates the spreading for all vehicle types from the road surface position and not from the actual source location. For receivers close to the road this error can change the results up to 1 db. Instead of calculation the spreading for the sources position (A) TNM 3.0 uses the road surface (B) as the source position for the spreading calculation for all vehicle types. Action needed: Calculate the spreading for the position of the source, not the road surface! Re-do the master values for the consistency test. TNM 3.0 problems Page 1

2 Problem #2 : Dissection of the road All formulas pertaining to line sources first need to break down a line source to one or multiple point sources depending on the geometrical situation, the length of the line source and distance between the source and the receiver. TNM is converting the line source using an angular method. The road is divided so that each segment does not exceed an angular width of 10 degrees. For multiple reasons this opening is too coarse and will cause uncertainties exceeding the envisioned accuracy of 0.25 db. The problem occurs at the end of the source line and in places where the road is partially shielded. Dissecting the sources should be done in a way where inserting extra coordinates in the road and thus forcing a finer resolution, should not cause differences. We found that the dissection caused deviations of up to 1 db. Dissecting the road in segments should not be done like in the left sketch below but like in the one on the right. If the ratio between the size of the segment and the distance from the middle of the segment to the receiver exceeds a certain value, the road needs to be divided and the checking needs to be done again until all segments have a ration that is bigger that the threshold value. A value of 4 seems to be appropriate. For practical reasons dissection should also been stopped if the length of segment is below a certain value (e.g. 1 m). Dissection is not only necessary for the geometrical spreading but is rather important for ground effect and screening. When a road segment is intersected at a shallow angle, it is not enough to regard a segment of the road as it fits into a 10 degree slot, the segment needs to also have the ratio of length of the segment versus distance from the source point observed. Consider a road segment of a kilometer in length and the receiver point 10 meters away from the road. What would be the distance for the spreading? TNM 3.0 problems Page 2

3 Another problem associated with a fixed angular opening for a road segment to be considered as a point source is the question where the effect of changing ground from soft to hard be assessed reliably. Here some examples: TNM 3.0 problems Page 3

4 It can be argued that uncertainties as were seen in the previous pictures do not occur in a live example because a road of 300 meters normally would not be entered with just 2 or 3 coordinates. One thing is absolutely certain, it should not matter where the user divides the road, the results should not depend on this. If we take the example in the last pictures and divide the road so that each road segment is just one meter long, this should give us the most accurate prediction. It does! The next picture shows the outcome of this idea. Calculated with steps of 10 degrees but with road segments of one meter of length, the road looks like the example in the SoundPLAN 8.0 case with a full projection. The uncertainty that for bigger distances was in excess of 5 db is gone. This example can easily be repeated with the original TNM. Have fun experimenting! TNM 3.0 problems Page 4

5 The projection method not only applied to obstacles but also to ground type and mitigation areas is clearly superior to projecting at obstacles but leaving a fixed scanning angle of 10 degrees. The angular incremented pictures were created with SoundPLAN 6.5 varying the angular increments, the picture with full projection was created using SoundPLAN 8.0. It is clear that 10 degrees as the standard angular scanning is not sufficient! Action needed: Dynamically check and reduce the ratio of length of the segment / distance source to receiver and reduce the size of the scanning angle to at least 1 degree! Re-do the master values for the consistency test. Problem #3 : Receivers in the extended roadway The consistency test suite in section has a systematic problem recognizing hard ground close to the end of a road segment. When receivers are positioned in an angle to the road segment where the linear connection between the road source and the receiver is no longer intersecting with the extended edge of road line, TNM 3.0 does not recognize that the roadway is hard ground and calculates the ground effect with soft ground only. Here some examples: A line source without directivity, line source and end of line with the road as source A road source would look a little bit different than the line source in the example above as the road source has an implicit directivity associated with the line source and the ground is not the same for the entire example. The roadway is hard ground and the terrain around the road is soft ground. For this reason the example 2 below shows on the left side that the noise from cars and trucks carries farther if the transmission has a lot of the reflective road surface in its path instead of the more absorbent soft ground. For this reason the contour lines in the example on the left experience a wider and wider spacing in the direct path of the road. The example on the left was calculated with the TNM 3.0 adaptation in SoundPLAN, the right example was calculated with the original TNM 3.0 software where we extracted the receivers from the grid of receivers and made them stand alone receivers. We transferred the receiver coordinated to TNM 3.0, did the noise calculation and then transferred the results back into a grid to be mapped by the SoundPLAN graphics. TNM 3.0 problems Page 5

6 TNM 3.0 calculations with SoundPLAN and with TNM 3.0 original. Why are there differences between the two versions, and why does it matter? The TNM 3.0 calculation contains a systematic error for cases where the transmission from source to receiver is not passing over the outer edge of the road segment. Below see the sequence in pictograms. TNM 3.0 problems Page 6

7 Apparently TNM 3.0 needs to intersect with the edge of the extended road segment in order to understand that the ground between the source and the receiver is partly hard ground. The diagram from the TNM Consistency Test Suite shows the inconsistency, the road is 12 ft wide and all receivers less than 7 ft from the road center line have the same value. Between 6 and 7 ft the value jumps 4dB because all of the sudden the road edge was intersected and the transmission was partly on hard ground. Situations like the one described above are common for example with a T-junction of a road. It is not clear what will happen with the overlap of hard ground in such a case. When the Test Suite was generated the levels between the road axis and the extended edge of the road should have raised flags as there is no acoustical reason for this behavior and the directivity is definitive not the culprit. TNM 3.0 problems Page 7

8 In the picture below all receivers that are inside the extended road edge lines calculate the ground effect only for soft ground! For a single road segment the error results in noise levels being up to 3.6 db too low, however for normal noise calculations the error described above is not as important as it appears as there are multiple road segments active at the same time. Many of these errors, glitches and unwanted patterns are only detected when the noise modeling software permits noise contours to be calculated and drawn. When SoundPLAN introduced Grid Noise Maps over 20 years ago, we had to correct many of these effects and it can be assumed that TNM will have to go through the same phase once noise contours bring these uncertainties to life. Action needed: Correct the problem for ignoring hard ground on the extended roadway! Re-do the master values for the consistency test. Problem #4 : Overlapping Roads in TNM 3.0 A road is defined as a hard surface with noise sources associated with it. As many roads have multiple lanes, it should be possible to simulate a multiple lane road with multiple roads as long as the edge of the asphalt band is consistent. Aside of simulating the source characteristic of a road with the types and numbers of vehicles, all the software needs to do with the road is to establish where the hard surface ends and where the soft ground begins. In TNM 3.0 there seems to be a problem with this. Most of the consistency tests are using multiple parallel roads. In half of the sets the road width overlap slightly in the other half the overlap is significant. Unfortunately the results of TNM 3.0 calculations vary up to 1.5 db when the road width is varied in a way where the source position remains identical and the overall edge of the road also is constant. As this error is not only an error in the near field of the road, it is very important to correct. TNM 3.0 users will think that TNM will calculate the overall edge of the road correctly but it does not, thus for the user the outcome is a random error of up to 1.5 db. TNM 3.0 problems Page 8

9 Road #1 and road #2 are overlapping. If the width of road #2 is varied so that the edge of road string is still within road #2, the noise received at the receiver changes up to 1.5 db, the emitting road axis remains in the same place! It seems that the edge of road string from road #2 needs to be eliminated for the part where another road is extending the pavement without a gap. There is also some evidence that some road segments are ignored in the calculations when road segments overlap. For sure noise contour maps also would show Moiré effects starting from these overlap areas. Action needed: Calculate the edge of the pavement correctly regardless of the degree of overlap of the road! Re-do the master values for the consistency test. Additional problems? So far we have tested Test Suite positions 1..3, further progress is not possible until the errors are fixed. With further test cases building on the basic cases, we cannot determine where deviations come from, if they are caused by the problems described above or if we have a problem in SoundPLAN or if TNM has additional problems. As the corrections not only will change the outcome of the Consistency Test Suite but all future work, the software should be updated quickly and the data for the Consistency Test Suite redone. This definitively should be done before even a limited release of the TNM 3.0 software. If the written text of the standard is the valid definition of TNM, then neither TNM 2.5 nor TNM 3.0 would pass the Consistency Test. TNM 3.0 problems Page 9

10 Why are these problems important? SoundPLAN GmbH, the parent company to SoundPLAN International LLC has continuously developed noise modeling software for the past 30 years. Currently they are involved in the development of the new ISO 17534: Acoustics Software for the calculation of sound outdoors Part 1: Quality requirements and quality assurance. Naturally we want to apply this quest for quality to our entire product and as an extension to the standards we use. We are interested in making TNM 3.0 the best road noise modeling standard available and thus want to make sure that it is consistent and that the results are as good an interpretation of the physics as possible. We also have an interest that the results of TNM fit the propagation of other noise sources such as railways that sometimes need to be evaluated in combination with the road noise model. As TNM itself cannot supply a solution for this, SoundPLAN offers an alternative capable of many options from true optimization/minimization of noise barriers to various graphics options to present noise contours on the terrain or in a vertical cutting. SoundPLAN is used worldwide for environmental impact studies, the mitigation of noise from roads and railways, the noise impact on residential buildings, the noise exposure of workers in the factory environment, the noise the study of the health effects of noise and the mitigation of aircraft noise. The software is standards based with currently over 75 national and international standards implemented and verified, the 1978 FHWA model as well as TNM 2.5 and TNM 3.0 are implemented. TNM officially is used to assess the need for noise barriers along roads in Federal Aid Projects. In practice the software is used in a much broader way to assess noise from highways, sometimes stand alone, sometimes along with other noise sources such as railways. So far TNM does not have the capabilities to draw noise contour lines or to superimpose noise from roads with the noise from railway or industrial sources. TNM is currently available as TNM 2.5 with the up-coming version TNM 3.0 in preparation. SoundPLAN has implemented the rules of TNM 2.5 and is preparing/testing the rules for TNM 3.0. SoundPLAN users in the USA and also in Canada, Israel, New Zealand, Australia and Singapore are interested in using the SoundPLAN software for road noise with the TNM model and want to make sure that the results are accepted by the departments of transport (Federal, State and local). The Federal Highway Administration has generated a Consistency Test Suite and defined Acceptable Margin Answer Keys for the FHWA TNM. SoundPLAN has started implementing the test questions as a SoundPLAN project and checked the results against the official results. Most of the answers from SoundPLAN are within the margin of 0.25 db but some answers are not. For the cases where SoundPLAN deviates in excess of the Acceptable Margin, we started researching why. Some of the results are laid out in the text above, unfortunately our tests can only complete when the errors in TNM are corrected. TNM is a complex standard, it was the first to calculate ground effect and screening with a phase correct algorithm. SoundPLAN started implementing TNM 17 years ago, since then we implemented and tested the other phase correct model, the Scandinavian Nord Implementing and testing Nord 2000 required our software developers to closely cooperate with the developer of the standard. In the process, errors were found on both software interpretations and were reconciled. Comparing the outcome of two software products will always help to find systematic errors, unfortunately not only in the software to be tested but also in the master program. Random deviations are much more difficult to identify with single receiver results. Often the numerical results in themselves seem correct but if they are compared to a receiver close by they are showing differences TNM 3.0 problems Page 10

11 that are bigger than expected. In SoundPLAN we found many inconsistencies when we mapped noise contours, something that TNM has not experienced yet. Scanning roads with 10 degrees of angular increment is such a problem that will only manifest itself if a noise map is drawn. As it can be assumed that the future of TNM 3.0 also will involve noise contour lines, it is wise to correct these problems before the users point out these inconsistencies. It is our opinion that TNM 3.0 could be much improved if reflections from individual buildings would be added to the portfolio. SoundPLAN has implemented this feature and for more than 30 years SoundPLAN users are successfully using this feature to predict noise levels where reflections from individual buildings cannot be ignored. Driving along Interstate 5 through downtown Seattle would support this concept. For all their users, SoundPLAN is interested in earning the status of equivalency to TNM, being part of the noise community we also offer our help in debugging the TNM software. Having implemented and tested TNM and Nord 2000, we know how valuable it is to have a review of the software against a second set of programmer s eyes. To get the best product, we have offered our cooperation in the past and still do! We offer our cooperation in trouble shooting the calculation core of TNM 3.0, helping establish the correct results of TNM and the TNM Consistency Test Suite. As we have been continuously involved in leading the development of noise modeling software for over 30 years, we have the expertise to improve and optimize noise modeling software so that it becomes cutting edge again. With best regards Arne Berndt; arne@soundplan.com SoundPLAN International LLC 80 E Aspley Ln; Shelton, WA For an overview of SoundPLAN please click here: or have a look at our website: TNM is a trademark of the US Government TNM 3.0 problems Page 11