Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
TCRP D-12 Final Report 44 5. FIELD MEASUREMENTS 5.1 Measurement Overview The overall goal of the field measurements was to estimate the exposure of each survey respondent to ground-borne vibration and noise using a wide variety of exposure metrics. The approach taken was to focus on obtaining a sufficient number of responses to the social surveys so that valid statistical inferences could be drawn from the data. There were 1,306 respondents to the social survey; clearly it was not feasible to measure noise and vibration inside the residence of each respondent. Instead, measurements were made at 41 residences, and ground-surface measurements were made at an additional 100 locations (the system-by-system breakdown is given in Table 12). Grid measurements were interpolated to estimate vibration levels at each interview location based on its position along the alignment, and perpendicular distance from the alignment. The calculated exterior vibration levels were then adjusted to estimate the vibration and noise levels inside each residence, based on the exterior and interior measurements at the 41 residential measurement sites. At each system, the field measurements were performed over a three to five day period. For residential test locations, appointments were made allowing approximately one hour for the measurements plus sufficient time to set up and remove the equipment. The following general approach was used to select measurement sites, and to acquire and analyze the vibration and noise data. 1. Sensor Orientation: All vibration measurements were made with the sensor oriented in the vertical direction. Previous research has shown that the vertical vibration in residential structures tends to be the most severe, and also that people are most sensitive to vertical vibration. Field Measurements 2. Measurement Sites: Residential test locations were selected from the list of people who indicated in the telephone survey that they would be willing to have measurements made in their home. An attempt was made to select residences distributed over the study area. 3. Residential Measurements: At the appointed time, the measurement equipment was set up in the room where the resident indicated that the ground-borne noise and vibration was most noticeable. This was generally a living room, bedroom or basement family room. A vibration measurement was also made outside at the approximate setback distance from the alignment. 4. Grid Measurements: Measurements of ground-surface vibration were performed in a grid arrangement to characterize the ground vibrations over the entire survey area. The purpose of the grid measurements was to provide a means to estimate vibration exposure at all survey locations. The grid measurements were distributed throughout the study area and were typically made on street curbs or sidewalks next to the alignment. Outdoor residence measurement sites were also used as grid points for interpolation purposes. 5. Reference Position: During the residence and grid measurements, simultaneous vibration measurements were also made at one fixed reference position. The reference data provided a way to confirm that the trains that passed the residence/grid location were typical of the fleet.
TCRP D-12 Final Report 45 6. Perpendicular Measurements: At least one measurement series was performed at various distances along a perpendicular to the alignment. These data were used to develop distance attenuation relationships for each test area. 7. All vibration measurements were made using seismic accelerometers with a nominal sensitivity of 1 V/g or 10 V/g. Sound measurements were made using 12.7 mm (½â) microphones. The vibration and sound waveforms were recorded in the field using digital flash card recorders. 1. Initial Processing: The recorded data was processed using MATLAB to calculate a variety of sound and vibration metrics. Data Analysis 2. Indoor-Outdoor Levels: The residential measurements were used to characterize the relationships between indoor vibration, indoor sound and outdoor vibration. 3. Adjustments: Data pertaining to specific grid and residential locations were adjusted if the measured sample of trains was not typical of the fleet (as determined by the large number of measurements at the reference position). The following procedure was used to estimate the ground-borne vibration and noise levels inside each respondentâs residence: Vibration Predictions 1. Position along the alignment: The linear position of the residence along the alignment was calculated. The two closest grid points that bracketed the residential position were identified. 2. Distance attenuation: The distance of the residence from the alignment was calculated and the distance-attenuation relationships were used to adjust the levels at the bracketing grid points to this distance. The exterior ground vibration level at the residence was then estimated by linear interpolation of the distance-adjusted grid levels. All interpolations were made based on logarithmic (decibel) levels. 3. Outdoor to indoor: The indoor vibration levels were estimated based on an average outdoor-to-indoor adjustment. 4. Vibration to sound: The indoor vibration levels were used to estimate the indoor sound levels, based on an average adjustment determined from the measured residential data. Interpolation calculations were performed on a 1/3 octave band basis, resulting in estimates of indoor sound and vibration at each of the 1306 residences where telephone surveys were conducted. 5.2 Residential Measurement Location Selection The final interview question asked whether respondents would consider allowing measurements to be made inside their home. Homes suitable for vibration measurements were selected from respondents who answered âyesâ to this question. Attempts were made to select homes throughout the interviewing area, although this was not always possible given the limited number of people willing to permit access. Willing residents were contacted by telephone to arrange appointments. Reminder letters were mailed about two weeks before the planned test date, and follow-up telephone calls were made
TCRP D-12 Final Report 46 the week prior to the scheduled visit. Cooperation was generally very good. In the case of the few no-shows, it was possible to make measurements in a nearby neighbor's residence. Typically, three to four residential measurements were done per day, with grid measurements being performed between residential visits, where possible. The approximate latitude and longitude of the dwelling façade closest to the alignment was determined using aerial mapping software. Figures showing the location of each interview location may be found in Appendix A. For surface systems, the latitude and longitude of discrete points along the alignment centerline were determined through aerial mapping software. For sub- surface systems, the agencies provided the necessary plans and information to locate the alignment centerline. The alignments were digitized at 30 cm intervals and the digitized models were used to calculate 1) the distance from each residence to the alignment centerline, and 2) the position of the residence along the alignment (from a starting reference point usually just outside the test area). Figures showing the alignment in each test city may be found in Appendix A. In addition to the specific residential measurements, grid measurements were also made to determine the vibration gradients along, and perpendicular to, the alignment. The grid points were selected to provide an indication of the vibration distribution throughout the test area. Figure 14 shows the grid point distribution used for the New York test area. One measurement point per short block and two points per long block were used. Note that the blank area in the southeast of Figure 14 is a park. In addition to the grid points along the alignment, measurements were made along a line perpendicular to the alignment. The perpendicular measurements were used to calculate vibration attenuation with distance. Where appropriate, exterior residence measurements were used as additional grid points. The locations of grid points were identified according to visible landmarks (street intersections, buildings). Aerial mapping software was used to calculate latitude and longitude coordinates of the grid points. The grid locations for all test cities are shown in Appendix A. Figure 14: New York Grid Measurement Points
TCRP D-12 Final Report 47 5.3 Field Measurement Procedures 5.3.1 Instrumentation All data collected in the field was recorded using Rion Model DA-20, four-channel digital recorders. The DA-20 systems record data as WAV files on compact flash cards. Table 13 summarizes the instrumentation that was used during the study. All vibration data was acquired as acceleration. Table 13: Instrumentation Summary TYPE SENSOR / INSTRUMENT PREAMPLIFIER SIGNAL CONDITIONER Data Recorder Rion DA-20 n/a n/a Vibration Wilcoxon 731A accelerometer n/a Wilcoxon P31 Vibration PCB 393A03 accelerometer n/a PCB 480E09 (1 channel) PCB 480B21 (3 channel) Vibration Calibrator Hardy Instruments HI 823 n/a n/a Vibration Calibrator PCB Model 394C06 n/a n/a Sound B&K 4189 12.7 mm microphone PCB 426E01 PCB 480E09 Sound Norsonic 1225 12.7 mm microphone Norsonic Model 1201 n/a Acoustical Calibrator B&K 4220 Pistonphone n/a n/a 5.3.2 Measurement Approach Figure 15 shows a schematic of the field test configuration. The measurements consisted of: 1. Residential interior measurement: Sound and vibration were measured inside a normally occupied room that the resident identified as a location where the transit noise and vibration was most noticeable, usually a living room, bedroom, or basement family room. An accelerometer was affixed to the floor near the middle of the room, and a microphone on a tripod was placed near the accelerometer. The microphone was set at a height of about 150 cm (5 ft) above the floor. Figure 16 shows a typical residential measurement installation. The indoor residential measurements were designed to be completed in approximately one hour to minimize the inconvenience to the resident. Most measurements were performed during daytime hours. 2. Residential exterior measurement: Ground vibration was measured immediately outside the residence. Exterior sound was also measured when the tracks were at-grade in order to determine whether the indoor noise levels were confounded by airborne noise from trains. The exterior residential measurement point was generally at the approximate setback distance of the structure from the alignment. Figure 17 shows a typical exterior measurement. The alignment was to the right in this photograph. 3. Grid measurement: Grid measurements were typically made on street curbs or sidewalks next to the alignment. Interpolation of the grid data was used to estimate exterior vibration levels at each interview location. Figure 18 shows a photo of a typical grid measurement location.
TCRP D-12 Final Report 48 The data recorder is the blue instrument at the bottom of the photograph. The instrument in the middle of the photo is the accelerometer signal conditioner. 4. Perpendicular measurement: The perpendicular measurements were typically made on the curb of a cross street that was perpendicular to the alignment. The perpendicular data were used to estimate vibration attenuation with distance from the alignment. The perpendicular measurements generally consisted of four to five positions at distances up to 75 m from the alignment. 5. Reference position: Simultaneous measurements were made at one common reference location for each cluster of grid/residential measurements. A large number of train events were captured at the reference location, which made it possible to characterize the fleet, statistically. The data from the grid/residence locations (typically 10-20 trains) could then be checked to see that the sample was typical of the fleet; if not the data was adjusted accordingly. Figure 19 shows a reference location that was used in New York City. The sensor was on the curb and the recorder and signal conditioner were inside the vehicle. Because the vibrations from trains were usually most easily detected at the reference site, the member of the test crew at this location was responsible for logging train events. The log of event times simplified identification and confirmation of train passages during data post-processing. Figure 15: Configuration of Field Measurements
TCRP D-12 Final Report 49 Figure 16: Typical Residential Measurement Setup Figure 17: Typical Exterior Measurement Setup
TCRP D-12 Final Report 50 Figure 18: Typical Grid Measurement Setup Figure 19: Typical Reference Measurement Setup
TCRP D-12 Final Report 51 5.3.3 Field Log Sheets and Field Crew Field log sheets were prepared for each type of measurement; reference, residence interior, residence exterior and grid point. Samples of the log sheets may be found in Appendix C. The typical field crew consisted of three people who communicated via mobile phone. For a residential measurement, one person was stationed at the reference location, while the other two were responsible for measurements inside and outside the residence. For grid measurements, one person was stationed at the reference location while the other two went to separate grid locations. 5.4 Data Analysis Procedures 5.4.1 Initial Processing The overall goal of the measurements and analysis was to obtain useful descriptors of the vibration and noise environment inside each residence for correlation with the questionnaire responses. Many different descriptors were calculated because it was not known beforehand which one(s) would correlate best with questionnaire responses. Processing and storage of reduced field measurements were accomplished in a manner that made it possible to compute additional descriptors as necessary. The field recordings were analyzed using MATLAB. The steps in the analysis were: 1. The recorded calibration signals were analyzed to determine sensor scale factors. 2. Each data file was analyzed to calculate 1/3 octave band spectra at 250 ms intervals (RMS averaging). Acceleration spectra were converted to velocity in the frequency domain using the appropriate 1/3 octave center frequencies. All 1/3 octave band spectra were stored as ASCII text files. 3. The 250 ms data were analyzed to identify train events at interior, exterior, and reference locations. Although the recorder clocks were synchronized to within ±1 second, there was often a time shift when the reference position was not at the same track location as the grid/residence measurement (the train arrived at the reference before it arrived at the residence, for example). Any time-offsets were accounted for to ensure that the same train produced the sound/vibration at each of the measurement sites. Once the data records were synchronized, train passage events were identified using the field log sheets and by visual inspection of the vibration time histories. There were many examples where the events were evident only at the reference position because the residential levels were masked by the background sound/vibration. 4. Train passage events were verified by reviewing the vibration time histories and the 1/3 octave band spectra for each measurement site. Events that were corrupted by sound/vibration from other sources were removed. An example of corrupted vibration data is shown in Figure 20. 5. The 1/3 octave band spectra starting approximately 10 seconds before the event and continuing until approximately 10 seconds after the event were stored for subsequent calculation of noise and vibration metrics for each event, and for calculation of various exposure metrics at each site. One step that was not taken was to group the events. For example, at many sites it would have been possible to identify which track the train was using based on the frequency spectra and arrival times. For example, Figure 21 shows the data from a site where the inbound and outbound
TCRP D-12 Final Report 52 trains had clearly different vibration characteristics. Trains were not separated into groups because 1) residents usually would be unable to distinguish which track the train was using, and 2) because the concern for this project was the overall exposure to vibration, not the characteristics of individual events. Likewise, in cases where there was a mix of commuter rail and rail transit operating in the same right of way, no attempt was made to distinguish between the two. For two of the test systems, however, freight and transit operations shared the same right of way. In these cases, the vibration from freight trains was ignored because 1) the freight trains were so infrequent it was not practical to obtain measurements at all residential/grid sites within the time available, and 2) the number of survey respondents exposed to mixed freight and transit traffic was less than 30, which was insufficient to draw statistical inferences. Figure 20: Example of Event Screening for Corrupted Data The left plot shows the 1/3 octave band vibration levels measured inside a residence in Dallas for all of the apparent train passages. The gray dashed line shows the ambient (no trains) vibration spectrum and the heavy gray line shows the linear average of the train-induced spectra. The right plot shows the corresponding vibration time histories. The highlighted event (in red) does not appear to have been caused by train vibration. The abrupt increase and decrease in vibration at about 5 seconds clearly is inconsistent with the other train passages and was likely to have been caused by something inside the home. This event was subsequently removed from the analysis.
TCRP D-12 Final Report 53 Figure 21: Example of Direction- Dependent Vibration Spectra This is an example where vibration from inbound and outbound trains had consistently different characteristics. Trains on the closer track exhibited higher levels between 80 Hz and 100 Hz. The heavyweight black line is the linear average and the lightweight gray line is the background (no trains). The error bars are the average background plus and minus one standard deviation. Train events and spurious noise (e.g., people talking) were removed before calculating the average background level. This also is an example where the train vibration was well above the background vibration at the outdoor position, but the vibration and noise were only marginally above the background at the indoor position. 5.4.2 Secondary Processing The 250 ms 1/3 octave band data formed the basis for subsequent calculations of event metrics and summary metrics for each site. Once the train events had been identified, the 250 ms data were used to calculate the following metrics on a 1/3 octave band basis for each train event: Lmax: the maximum RMS vibration level using RMS 250 ms, 1 second and 5 second duration averages, Leq: the RMS average over the event duration defined by the 3 decibel down points from the 1 second Lmax, SEL: the total RMS energy over the event duration defined by the 10 decibel down points (from the 1 second Lmax), normalized to a 1 second period. Note that the terms Leq, SEL and Lmax describe the characteristics of the passby, and can be in terms of either sound or vibration.
TCRP D-12 Final Report 54 In addition, the duration of each event was calculated and stored. (The duration was defined as the time between the 3 dB down points that were used for calculating Leq.) The analysis frequency range extended from 5 Hz to 200 Hz (1/3 octave band center frequencies). To allow maximum flexibility, the 250 ms data for each train event were also stored so that additional metrics could be calculated. The Leq, SEL and Lmax values for each event were then averaged to obtain the following quantities that were used to calculate different measures of vibration and noise exposure: Average Leq: The linear average of the Leq decibel levels and the standard deviation, Average Lmax: The linear average of the Lmax decibel levels and the standard deviation, Average SEL: The linear average of the SEL decibel levels and the standard deviation, RMS Average SEL: The root mean square average of the event SEL values using the following equation,     ï£ ï£« Ã= = â = N i absAvgRMS SEL abs iSEL N SEL SEL 1 2 10_ 20 )(1log10 10 RMQ Average SEL: The root mean quad average of the event SEL values using the following equation,     ï£ ï£« Ã= â = N i absAvgRMQ iSELN SEL 1 4 10_ )( 1log5 Average Event Duration: The average time between the 3 dB down points. The effects of the different averaging methods (linear, RMS and RMQ) can best be shown with an example. Consider the five decibel levels of 70, 75, 80, 85 and 100. The averages using the three different averaging methods, with and without the 100 dB level, are: Averaging with 100 dB without 100 dB Linear: 82.0 dB 77.5 dB -4.5 RMS: 93.2 dB 80.6 dB -12.6 RMQ: 96.5 dB 82.2 dB -14.3 Difference The 100 dB level dominates the RMS and RMQ averages. Removing the 100 dB level reduces the RMS and RMQ averages by 13 and 14 decibels respectively but reduces the linear average by only 4.5 decibels. It is commonly expected that human annoyance with noise and vibration is associated with the loudest (highest vibration/noise) events and therefore an RMS or RMQ average will provide a better measure of potential annoyance than a linear average. The average Leq, Lmax, and SEL were then used to calculate the site metrics listed in Table 14. The metrics include the average vibration level and several measures of the 24-hour vibration exposure. For the Leq, Lmax and SEL velocity metrics, the mean level plus two standard deviations was also calculated. (Klaeboe (31) showed that the mean plus 1.8 standard deviation level (vw95) was a good approximation for the loudest trains in a fleet.)
TCRP D-12 Final Report 55 Table 14: Summary of Site-Level Measurement Metrics Measure Number of Related Measures Description Passby Duration 1 Based on 3 dB down points from 1 second Lmax. Leq, Velocity and Sound 20* Energy average based on 3 dB down points. Lmax, Velocity and Sound 20* Maximum using an RMS averaging time of 1 second. SEL, Velocity and Sound 20* Energy normalized to 1 second using the 10 dB down points. Velocity Exposure RMS 20* 24-hour exposure to vibration based on the sum of the RMS average SEL values. Velocity Exposure RMQ 20* 24-hour exposure to vibration based on the sum of the RMQ average SEL values. Acceleration 7** Acceleration calculated from the 1/3 octave band spectra of the Leq, Lmax and SEL. A-weighted Velocity and Sound 7** Overall level calculated from the 1/3 octave band spectra of the Leq, Lmax and SEL with A-weighting applied. Wm-weighted Velocity 7** Overall level calculated from the 1/3 octave band spectra of the Leq, Lmax and SEL velocity with Wm weighting applied. LFSL Velocity and Sound 7** LFSL (low-frequency sound level) calculated from the 1/3 octave band spectra of Leq, Lmax and SEL. * Overall plus each 1/3 octave band from 5 Hz to 200 Hz ** Overall level for Leq, Lmax, SEL (linear, RMS and RMQ averages), and exposure (RMS and RMQ) 5.4.3 Quality Assurance/Quality Control A substantial amount of noise and vibration data was collected during the field measurement phase of this project. A key component in the design of the data collection and analysis procedures was to ensure that the data obtained was generated by trains and not corrupted by background vibration and noise. The following steps were taken to ensure data quality: 1. The MATLAB routines for processing the WAV files were verified by comparing the results to those obtained independently by playback through a commercial spectrum analyzer. This validation of the MATLAB procedures was performed prior to fully processing the field test data. 2. Sound and acceleration calibration signals were recorded in the field using acoustical calibrators for the microphones and portable vibration shakers for the accelerometers. The calibration levels were verified prior to processing the data files. 3. The field log sheets were scanned and saved as PDF files so they could be readily referenced during the data analysis process. 4. Simultaneous measurements were made at a reference position during all of the grid/residential measurements. The times when trains passed the reference site were noted, which helped to identify which vibration events were caused by trains, and which were caused by other vibration sources. 5. Both the time history and the 1/3 octave band spectra for each possible train event were inspected for each measurement position. Events that were clearly not caused by trains were
TCRP D-12 Final Report 56 eliminated from the analysis. Note that trains that caused unusually high vibration and/or noise levels were retained in the analysis because these trains were part of the residentsâ vibration exposure. One useful way to determine whether a vibration event was caused by a train was to verify that the event occurred simultaneously at each of the measurement positions (interior, exterior, reference). 6. The background vibration (no trains) was also calculated at each site. The background was determined from the linear average vibration level (in decibels) with train events excluded as well as any unrepresentative vibration events, such as people walking in the building. Displaying the background vibration along with the train-related spectra helped to identify spurious events. Since the data was recorded, it was also possible to listen to the appropriate section of original WAV file to verify that the vibration was produced by a train and not something else. 5.5 Measurement Overview Although this study was focused on determining human response to building vibration generated by rail transit operations, the field measurements represent a valuable body of information on the characteristics of the ground-borne vibration and noise found at five North American rail transit systems. The measurement results have potential application to the methods that are currently used to predict ground-borne vibration from proposed new transit systems, and when responding to complaints about ground-borne vibration and/or noise. The following sections present some important details of the measurement results including 1) the average relationships between outdoor vibration and indoor vibration/noise, 2) the relationships between several different metrics for characterizing ground-borne vibration/noise, and 3) the system-to-system variation in the observed vibration levels. 5.5.1 Relationships between Indoor Sound, Indoor Vibration and Outdoor Vibration The average overall vibration and noise levels measured at each of the residential test sites are summarized in Table 15. The values in the table represent average RMS levels, based on the 3 dB down points (referred to as Leq in this report). Table 15 shows the average levels from the three standard residential measurement positions: 1) outdoor vibration, 2) indoor vibration, and 3) indoor sound. For all three measurement locations, the un-weighted overall levels and the A- weighted levels are shown. The calculation of the overall levels was limited to the 1/3 octave bands from 5 Hz to 200 Hz. The uppermost plot in Figure 22 shows the relationship between indoor vibration and outdoor vibration. Although there is considerable spread in the data (up to ±10 decibels), the best-fit line is very nearly given by, Lv(Indoors) = Lv(Outdoors), which means that, on average, the vibration level inside the residence was equal to the vibration level outside the residence. The relationship between indoor and outdoor vibration is determined by the coupling loss at the foundation/soil interface, the attenuation as vibration propagates from the building foundation into occupied spaces, and building structural resonances that amplify the vibration (floor resonances in particular). The data in Figure 22 suggests that these effects tend to cancel out for the residences tested in the D-12 study and, on average, the indoor vibration equals the outdoor vibration. This applies to single story slab-on-grade residences in Sacramento and
TCRP D-12 Final Report 57 Dallas, brownstones in New York and Boston, and two story duplexes with full basements in Toronto and Boston. The middle two graphs in Figure 22 show the relationships between indoor sound and outdoor vibration, and indoor sound and indoor vibration, respectively (all with no weighting applied). The lower two graphs show the same data with A-weighting applied to both vibration and sound. The best-fit linear curve fits and the correlation coefficients (R2) are also shown on the figures. In predicting the structure-borne sound in a room, it is commonly assumed that the sound pressure level is equal to the vibration velocity level of a representative surface (in VdB). Figure 22, however, shows that Lp = Lv is not necessarily a good model for the D-12 data. In fact, the present study suggests that Lp = Lv â 5 dB is a better model for the relationship between sound pressure and floor vibration. Some additional observations from the data summarized in Table 15 and Figure 22 are: ⢠The outdoor vibration level was nearly as good a predictor of indoor sound level as was the indoor vibration level. ⢠The correlation between sound and vibration improved when both were A-weighted. ⢠The majority of the locations where the indoor A-weighted noise levels exceeded 40 dB were in Toronto. The indoor A-weighted noise level exceeded 40 dB at eight of the 11 residences measured in Toronto. Of all the other residences in Boston, Dallas, New York, and Sacramento, in only one residence in Boston and one residence in Sacramento did the indoor A-weighted sound level exceed 40 dB. ⢠The difference between indoor sound level and indoor vibration velocity level increased as the sound level increased. For example, referring to the best-fit line in the lower two graphs in Figure 22, the average difference between the vibration level and the sound level was 2 dB when the A-weighted vibration level was 30 dB, and 9 dB when the A-weighted vibration level was 50 dB. This could be related to the influence of the ambient sound/vibration when the trains were reasonably quiet. ⢠Forcing the slope of the best-fit curves of A-weighted sound level as a function of A- weighted vibration level to be unity, the relationship is approximately: LPAwt = LVAwt - 5
TCRP D-12 Final Report 58 Table 15: Overall Vibration and Noise Levels Measured at Residences Transit System Survey Label Dist. from Track CL (m) Measured Overall Levels1,2 (dB) Outdoor Vibration Indoor Vibration Indoor Sound Linear A-Weight Linear A-Weight Linear A-Weight3 Toronto 10086 18.6 75.4 51.2 74.0 50.4 68.6 44.0 10124 16.8 77.3 51.1 78.1 55.3 69.4 43.1 10126 15.2 71.6 43.4 74.1 43.9 72.3 43.1 10230 5.8 74.8 51.9 82.2 64.5 69.7 52.7 20073 27.7 73.3 47.9 63.9 42.0 63.4 42.0 20088 19.8 70.1 50.6 76.4 54.1 64.5 42.5 20012 3.4 78.4 48.9 82.4 52.1 67.2 40.8 Opp4 15.5 80.8 56.9 73.6 49.1 67.8 43.8 70053 39.6 66.2 35.7 67.7 39.0 65.3 31.1 70068 83.5 55.8 32.3 61.6 39.5 56.8 37.3 70079 39.3 64.7 38.0 68.1 39.7 57.0 30.1 70106 19.8 76.7 50.6 79.7 48.6 70.5 41.1 New York Opp4 16.2 65.3 35.9 61.0 30.2 64.5 35.5 00040 16.2 66.5 40.1 75.3 40.8 71.9 28.8 00057 63.7 66.0 35.5 65.4 23.4 57.4 19.2 00110 16.2 64.9 33.6 70.1 28.0 72.6 30.9 00116 19.8 71.3 42.2 79.4 49.2 67.2 37.1 00207 14.0 67.5 38.7 69.9 33.0 65.4 27.8 00240 18.9 74.3 47.6 67.9 37.4 66.8 34.8 00254 26.8 74.4 43.9 70.6 41.3 65.4 33.3 00269 19.5 74.5 44.0 80.2 48.7 67.4 37.2 00272 56.7 61.1 31.0 68.4 40.5 64.6 33.2 Boston 20031 83.2 53.7 13.3 50.6 7.4 51.4 -1.1 20078 83.2 51.3 6.7 57.2 12.0 57.4 7.5 70134 57.3 62.1 7.2 58.2 2.1 67.5 22.4 70148 82.0 55.1 19.5 56.1 12.0 51.4 13.3 20000 15.2 70.1 37.6 80.8 50.8 70.3 43.0 70007 32.0 66.0 35.7 68.0 35.4 62.6 27.7 70080 103.0 63.2 33.5 63.8 33.2 51.6 19.5 Dallas 00053 89.0 43.6 13.5 46.2 5.6 55.4 20.7 00061 16.5 58.2 23.3 65.1 25.2 63.0 29.3 00065 11.6 59.3 29.9 53.3 26.4 --5 --5 Sacramento 00003 11.3 60.1 18.2 56.8 16.9 --5 --5 00010 11.6 43.5 8.3 46.7 9.0 45.8 6.5 00011 16.5 74.5 42.8 69.4 36.7 61.9 35.4 00026 89.6 69.6 36.1 67.4 35.2 64.6 45.5 00029 16.8 73.3 39.1 69.4 36.0 80.4 32.5 1. Levels are train passby Leq. The Lmax (1 second duration) levels were 1 to 2 decibels higher. 2. Overall levels for 5 to 200 Hz 1/3 octave bands. 3. Levels in bold exceed an A-weighted level of 40 dB. 4. Test of opportunity, occupant of intended test house was not home. 5. Indoor sound measurements were not performed at these residences because resident was not home. The indoor vibration measurement was performed on an open porch.
TCRP D-12 Final Report 59 Figure 22: Relationships between Indoor and Outdoor Measurements
TCRP D-12 Final Report 60 5.5.2 Relationships between Metrics As discussed earlier, a number of different measures have been used or considered for characterizing human response to ground-borne vibration and noise generated by rail transit operations. A goal of this study was to investigate which measures were best correlated with human response to the vibration/noise. The two basic categories of metrics are 1) metrics based on average train vibration levels and, 2) metrics based on cumulative measures of exposure to vibration. These metrics include a number of different weighting curves (see Figure 1 and Figure 2), most of which are equivalent to vibration velocity level over the 16 Hz to 80 Hz frequency range. The steps in the data analysis procedure consisted of analyzing the measurements, calculating average and exposure levels at each measurement position on a 1/3 octave band basis, and then using interpolation to estimate the outdoor vibration levels at each respondentâs residence. The different metrics were then calculated using the interpolated 1/3 octave band levels. As a first step in investigating the relationships between different metrics, a factor analysis was performed on the interpolated data. A factor analysis is a multivariate statistical technique that seeks to create a small number of analytic âdimensionsâ that identify subsets of variables that are closely related to one another, but not to other subsets of variables. The practical utility of a factor analytic solution depends on the physical interpretability of the resulting dimensions. If the various dimensions with which variables are associated cannot be readily described in useful terms, the factor-analytic solution is of no assistance in distinguishing among variables. Metrics used in the factor analysis consisted of the different measures of overall vibration level, cumulative exposure metrics, and the levels in each 1/3 octave band. The factor analysis was unable to identify physically interpretable subsets of vibration metrics that were both closely related to one another and at the same time different from other subsets. The very high correlation among measures implies that, for purposes of predicting community response to train-induced vibration levels in homes, custom, convenience, and cost are as reasonable a basis for selecting a vibration metric as any statistically-derived criterion. This observation is specific to the data collected for this study and may or may not be applicable to systems with substantially different vibration characteristics. However, it is clear in the present data set that any number of metrics would be equally effective at predicting annoyance. Since any metric was essentially as good as any other at predicting community response, it seemed prudent to select a series of metrics for the detailed dosage-response analysis that were familiar to the noise and vibration community and did not require specialized instrumentation. An effort was also made to the extent possible to choose variables for detailed analysis that were less-well correlated with one other. The metrics considered were: Overall Velocity: Most common methods for characterizing passby train vibration are equivalent to either the RMS average vibration velocity over the duration of the passby, or the maximum vibration velocity level measured during the passby. Figure 23 shows the relationship between the RMS average over the 3 decibel down points (Leq) and the Lmax using a 1 second averaging time. As can be seen in Figure 23, the two measures are highly correlated and, consequently, either would be equally effective at predicting annoyance.
TCRP D-12 Final Report 61 Figure 23: Relationship between Site Average Outdoor Lmax and Leq Lmax is the maximum passby level using a 1 second RMS average. Leq is average level over the 3 dB down points. Weighted Velocity: Figure 24 shows the relationships between un-weighted and weighted vibration velocity level using representative weighting curves. The figure clearly shows that these metrics are highly correlated with each other. Consequently, any of the weighted velocity levels would be as good a predictor of annoyance as any other. Acceleration: Although few standards for ground-borne vibration are defined in terms of the vibration acceleration level, it was considered in this study because it has a different spectrum weighting than velocity (acceleration places more emphasis on high-frequency vibrations). Figure 25 shows the relationship between Leq velocity (average between 3 dB down points) and Leq acceleration. From the figure it can be seen that even when using these two relatively dissimilar measures, the correlation is quite high (R2=0.91). A-Weighting: A-weighting is relevant because current standards for ground-borne noise are typically expressed in terms of the indoor A-weighted sound level. It is commonly assumed that the sound pressure level inside a room is proportional to the vibration velocity level of the vibrating room surfaces. Therefore, A-weighted vibration velocity should also be a relatively good predictor of A-weighted ground-borne noise. Figure 26 shows the relationship between the A-weighted velocity level and the un-weighted velocity level. The correlation between the two metrics is high (R2 = 0.80), although not as high as the correlation between un-weighted Leq and Lmax. Variables that are not well-correlated are desirable because they can be tested in the dosage-response analysis with the hope that one may be a better predictor of annoyance than the other. If the variables are highly correlated, by definition they will account for the same percentage of the observed variance.
TCRP D-12 Final Report 62 Figure 24: Relationship between Un-weighted and Weighted Velocity Levels Both graphs compare weighted vibration velocity to the un-weighted velocity. The left plot uses ANSI weighting, which rolls off at frequencies lower than 8 Hz. The right plot shows Wk weighted velocity. Figure 25: Relationship between Leq Acceleration and Velocity Levels
TCRP D-12 Final Report 63 Figure 26: Relationship between A-weighted and Un-weighted Velocity Level Measures of Exposure: A potential shortcoming associated with most current ground-borne vibration standards is they are based on the absolute train vibration level, and have little ability to account for the number of events per day, or the duration of a typical event. It seems reasonable that longer trains would be more annoying than shorter trains, and that annoyance would increase with the frequency of service in general, and nighttime service in particular. Table 11 summarizes the relative levels of exposure for the five transit systems tested in the D-12 study. The table shows that when the maximum levels of train vibration were the same, there was approximately an 11 dB difference between the exposure in Toronto (the greatest exposure) and the exposure at the light rail lines in Sacramento and Dallas. Table 16 summarizes the average event duration for the five test systems. The average durations were consistent with the typical numbers of cars in the consist (see Table 11). Table 16: Average Event Duration Transit System Duration, seconds1 Average Standard Deviation Minimum Maximum Toronto 9.4 1.6 5.2 13.5 New York 14.6 3.1 8.8 19.7 Boston 9.0 2.8 5.8 16.9 Dallas 4.0 0.8 3.0 5.7 Sacramento 5.5 1.6 3.4 10.1 1. Average time between 3 dB down points.
TCRP D-12 Final Report 64 The exposure metrics most familiar to the acoustics community are based on the equivalent sound level (Leq) over a period of time, most commonly 24 hours or one hour. For example, many impact criteria for noise are based on the Day-Night Average Sound Level (abbreviated as Ldn or DNL). Ldn is a 24-hour Leq with a +10 decibel penalty applied during nighttime hours to account for the increased sensitivity of people at night. As shown in Table 11, the ratio of daytime-to-nighttime trains did not vary significantly among the five transit systems. This is not surprising since, presumably transit agencies plan service around the conventional work day. Because there is no real variation in the day/night exposure metrics, it is not possible to gain any insight into the relative importance of daytime versus nighttime exposure to vibration. Consequently, the exposure metrics used to investigate dosage-response relationships do not include nighttime adjustments. RMQ is a metric that the British have proposed to characterize exposure. As discussed earlier, laboratory studies by Howarth and Griffin (8) indicate the following relationship between annoyance, number of events (N), and vibration amplitude (V): annoyanceNV â7.3 This relationship was used in developing the fourth-power vibration dose value (VDV) used in the British Standard (6), which is essentially an RMQ exposure metric. To calculate VDV as defined in the British Standard would have required analyzing the raw vibration data to obtain RMQ averages along with RMS averages. Although this would have been possible using MATLAB, the potential benefits of the additional processing were not sufficient to justify the considerable effort that would have been required to re-analyze the data. Rather, an approximate method was used to investigate the correlation between RMS and RMQ exposure using average SEL at each exterior measurement site. Figure 27 shows the relationship between the average SEL using RMS velocity and the average SEL using RMQ velocity. The two metrics are highly correlated (R2=0.986) with only a few locations where the variance from the best fit line is more than 1 dB. The average difference between the RMQ and RMS averages is approximately 1 dB. It was decided that the computational effort in calculating VDV did not justify the small potential benefit in accounting for slightly more of the observed variance.
TCRP D-12 Final Report 65 Figure 27: Comparison of RMQ and RMS Exposure Comparison of site average SEL calculated using RMS and RMQ averaging for all exterior measurement positions. SEL is the total energy of the event normalized to 1 second. 5.5.3 Vibration Spectra The frequency character of the vibration can be equally as important as the absolute level. This section presents an overview of the measurement results in terms of the 1/3 octave band spectra. The outdoor vibration levels are used for the comparison and the metric used is the RMS average over the 3 dB down points (Leq). 5.5.3.1 Comparison of Vibration Spectral Shapes Figure 28 through Figure 33 show the average vibration spectra obtained at each of the outdoor measurements locations (residence plus grid), at each of the transit systems. The upper plot in each figure shows the measured vibration spectrum at each site, along with a line corresponding to 72 VdB for comparison. The lower plot shows the spectra after being normalized such that the overall vibration level is 72 VdB. The purpose of showing the normalized spectra is to highlight the frequency characteristics of each system. The following observations can be made regarding the data shown in Figure 28 through Figure 33: ⢠The Toronto data have a distinct peak at 50 Hz that is evident in all but a few of the vibration spectra (Figure 28). None of the other systems exhibited such distinct peaks, particularly ones that that were so consistent. The fact that this peak shows up in the results for most of
TCRP D-12 Final Report 66 the measurement sites could be an indication that it is a characteristic of the Toronto situation (track, tunnel, rolling stock, soil conditions). ⢠A second, less distinct, peak shows up in the Toronto data in the 80 Hz to 100 Hz range. Both of the Toronto peaks are in the audible frequency range, and it is likely that these peaks are perceived inside buildings as audible noise. The absolute Toronto levels also exceeded 72 VdB at a number of locations, and it is expected that these would be perceptible by most people. The combined effect of audible noise and perceptible vibration could contribute to a higher level of annoyance in Toronto than at other systems where noise was more prevalent. ⢠The vibration spectra measured in New York (Figure 29) had a consistent shape with a broad peak in the 40 Hz to 80 Hz range. ⢠The maximum spectral levels near the Boston Blue Line (Figure 30) typically occurred in the 40 Hz to 60 Hz range. At the Blue Line, the vibration levels above 30 Hz decreased more rapidly with increased distance than did the vibrations below 30 Hz. This effect causes a low-frequency distortion that is evident in the normalized data. ⢠The Boston Back Bay spectra (Figure 31) fall into two distinct groups; Group 1 with maximum levels in the 60 Hz to 100 Hz range, and Group 2 that peaks around 20 Hz and then rolls off at higher frequencies. Further investigation showed that most of the Group 1 sites were located to the south of the rail right of way. The right of way is shared by commuter rail and Orange Line rapid transit trains. The Orange Line, which is located along the north side of the right of way, has a floating slab trackbed. It can be inferred from the measurements that the floating slab is effectively attenuating vibration at frequencies greater than 20 Hz. ⢠The vibration spectra at the Dallas sites (Figure 32) tended to be broadband, with maximum levels in the 30 Hz to 80 Hz range. The spectral levels were relatively low compared to the other transit systems. ⢠The vibration spectra from the Sacramento sites (Figure 33) generally had maximum levels in the 30 Hz to 60 Hz range. An interesting characteristic of the Sacramento data is a peak at 20 Hz that shows up in the results at some of the sites. Because the 20 Hz vibration levels are relatively low at the sites where this peak occurred, it does not have much influence on the overall vibration level, however, it is quite prominent in the normalized spectra. The Sacramento residence with the greatest measured vibration level was located near a crossover. In Figure 33 this corresponds to the red curve with solid red diamonds. ⢠Note that few of the measured outdoor vibration levels shown in Figure 28 through Figure 33 would exceed 72 VdB in any 1/3 octave band. There were a few sites in Toronto and New York, three at the Boston Blue Line, one in Sacramento, and none at either Boston Back Bay or Dallas. This indicates that most of the locations where vibration measurements were performed during the D-12 study would not be a considered to be âimpactedâ according to feelable vibration criteria used in the United States. Of course, this does not mean the sites would not be impacted based on ground-borne noise criteria. Figure 34 shows the energy averaged outdoor spectra measured at each transit system (sometimes referred to as âdecibel averagingâ). The top graph in Figure 34 shows the average spectra while the bottom graph shows the spectra normalized to an overall level of 72 VdB. The Toronto average spectrum is the only one that has appreciable energy above 60 Hz. It is also
TCRP D-12 Final Report 67 clear that Toronto is the only system where the maximum level occurs at a distinct peak in the vibration spectrum. As discussed above, the 50 Hz peak in the Toronto vibration spectrum occurred at most of the measurement locations. Based on the average curves and the current criteria for ground-borne vibration, Toronto would be expected to have the highest level of community annoyance closely followed by the Boston Blue Line. On a 1/3 octave band basis, the maximum vibration levels in New York were about 5 decibels lower than those in Toronto, indicating that a lower level of annoyance would be expected in New York than in Toronto. The vibration levels at residences in the Boston Back Bay, Dallas, and Sacramento were generally sufficiently low that limited levels of community annoyance would be expected, again based solely on a feelable vibration criterion.
TCRP D-12 Final Report 68 Figure 28: Average Vibration Velocity Spectra, Toronto Outdoor Measurement Sites
TCRP D-12 Final Report 69 Figure 29: Average Vibration Velocity Spectra, New York Outdoor Measurement Sites
TCRP D-12 Final Report 70 Figure 30: Average Vibration Velocity Spectra, Boston Blue Line Outdoor Sites
TCRP D-12 Final Report 71 Figure 31: Average Vibration Velocity Spectra, Boston Back Bay Outdoor Sites
TCRP D-12 Final Report 72 Figure 32: Average Vibration Velocity Spectra, Dallas Outdoor Measurement Sites
TCRP D-12 Final Report 73 Figure 33: Average Vibration Velocity Spectra, Sacramento Outdoor Measurement Sites
TCRP D-12 Final Report 74 Figure 34: Average Vibration Velocity Spectra, Outdoor Measurement Sites
TCRP D-12 Final Report 75 5.5.3.2 Outdoor to Indoor Adjustments At each of the residential test locations, the vibrations were measured simultaneously inside and outside of the residence. Noise was also measured inside the residence. Figure 35 through Figure 39 show the measured differences in the 1/3 octave band, indoor and outdoor spectra measured at each site. The RMS average vibration and sound levels over the 3 dB down points (Leq) were used for this analysis. The differences were calculated for each train event at each site and then presented as a site average. The upper plot in each figure shows the difference between the indoor vibration level and the outdoor vibration level (indoor â outdoor). The middle plot shows the difference between indoor sound and indoor vibration and the lower plot shows the difference between indoor sound and outdoor vibration. In the presentation of the data, any 1/3 octave band levels that were not at least two standard deviations greater than the background were excluded to avoid skewing due to the background levels. Figure 40 shows the average differences for all of the test sites. Figure 41 shows the average differences for each transit system, along with the overall average (all systems combined). Some observations and conclusions that can be drawn from these figures are: ⢠There is a large house-to-house variation in the D-12 data. In most cases the standard deviation is around 5 dB, with a tendency to increase with frequency. ⢠The average difference between indoor and outdoor vibration was nearly zero. The average for Toronto is within ±2 dB of zero over the entire frequency range. The New York average decreased with frequency at a rate of approximately 3 decibels per octave. (There were too few data points from Boston, Dallas and Sacramento to draw conclusions.) ⢠When all indoor vibration minus outdoor vibration data were combined (Figure 40), the average difference was effectively zero over the entire frequency range. ⢠Because the average indoor vibration was approximately equal to the average outdoor vibration, the curves of indoor sound minus indoor vibration and indoor sound minus outdoor vibration are similar. ⢠There is a consistent pattern of indoor sound being 3 to 5 decibels lower than indoor vibration over a frequency range from 31.5 Hz to 100 Hz. The differences are somewhat less for higher and lower frequencies. The 3 to 5 dB difference is consistent with the observed difference between A-weighted indoor/outdoor vibration and A-weighted indoor sound which was about 5 dB (see Section 5.5.1). ⢠There are no clear patterns based on the type of residence or the transit system. Almost all of the test residences were wood-frame construction. The residences in Toronto were primarily two story duplexes with basements, the residences in New York were three to five story brownstones and apartment buildings, the residences in Dallas and Sacramento tended to be single story slab-on-grade construction, and the residences in Boston were primarily two story duplexes, four to five story brownstones, or apartment buildings. ⢠An important factor to recognize in considering these results is that the indoor sound levels measured during train events were often near the background sound levels. In fact, only in Toronto did the indoor noise levels consistently exceed the background.
TCRP D-12 Final Report 76 Figure 42 shows the combined average curves obtained using different acceptance criteria in terms of the amount that a data point must exceed the background to be included in the average. The acceptance criteria range from no data points being excluded, to excluding all data points that do not exceed the average background level by at least three standard deviations. The following observations follow from this analysis: ⢠Indoor vibration minus outdoor vibration: The data acceptance criteria have virtually no effect on the average. ⢠Indoor sound minus indoor vibration: As the threshold for accepting data increases, the average difference drops. This effect increases with frequency. ⢠Indoor sound minus outdoor vibration: Because the average difference between indoor vibration and outdoor vibration is close to zero over the entire frequency range, the curves of indoor sound minus indoor vibration and indoor sound minus outdoor vibration are similar. The acceptance threshold has less of an effect on the indoor sound minus outdoor vibration than on the indoor sound minus indoor vibration. From the outdoor-to-indoor analysis it can be concluded that, for the D-12 data, the indoor vibration spectrum is effectively equal to the outdoor spectrum. Consequently, the exterior vibration levels are a very good predictor of the indoor levels, on average. While the indoor and outdoor vibration levels were comparable on average Figure 43 , there was a considerable variation when comparing specific houses. To illustrate this, consider three seemingly identical, adjacent brownstone buildings in New York. shows the average indoor vibration and sound spectra that were measured in each of the houses. The measurements were all made in the living room which faced the street (and the subway). The indoor sound levels were very similar, yet the vibration in one of the brownstones was 15 dB lower than was observed in the other two. The difference in vibration was likely due to a unique structural element or some other hidden peculiarity of the sensor position, however, the important thing to note is that there was no way to know by inspection of the exteriors which of the residences would have lower, or higher vibration levels.
TCRP D-12 Final Report 77 Figure 35: Average Outdoor-to-Indoor Differences, Toronto Solid blue line is average of all sites, dashed blue lines are ± 1 standard deviation. Only levels that were more than 2 standard deviations greater than background levels were included in the calculation of average differences.
TCRP D-12 Final Report 78 Figure 36: Average Outdoor-to-Indoor Differences, New York Solid blue line is average of all sites, dashed blue lines are ± 1 standard deviation. Only levels that were more than 2 standard deviations greater than background levels were included in the calculation of average differences.
TCRP D-12 Final Report 79 Figure 37: Average Outdoor-to-Indoor Differences, Boston Solid blue line is average of all sites, dashed blue lines are ± 1 standard deviation. Only levels that were more than 2 standard deviations greater than background levels were included in the calculation of average differences.
TCRP D-12 Final Report 80 Figure 38: Average Outdoor-to-Indoor Differences, Dallas Solid blue line is average of all sites, dashed blue lines are ± 1 standard deviation. Only levels that were more than 2 standard deviations greater than background levels were included in the calculation of average differences.
TCRP D-12 Final Report 81 Figure 39: Average Outdoor-to-Indoor Differences, Sacramento Solid blue line is average of all sites, dashed blue lines are ± 1 standard deviation. Only levels that were more than 2 standard deviations greater than background levels were included in the calculation of average differences.
TCRP D-12 Final Report 82 Figure 40: Average Outdoor-to-Indoor Differences, All Sites Solid blue line is average of all sites, dashed blue lines are ± 1 standard deviation. Only levels that were more than 2 standard deviations greater than background levels are included in the calculation of average differences.
TCRP D-12 Final Report 83 Figure 41: Average Outdoor-to-Indoor Differences for Each Transit System
TCRP D-12 Final Report 84 Figure 42: Comparison of Average Outdoor-to-Indoor Differences, using Different Data Thresholds âN x Stdevâ indicates that only levels that exceeded the background level by at least N times the standard deviation were included in the average.
TCRP D-12 Final Report 85 Figure 43: Comparison of Indoor Vibration and Sound in Three New York Residences The three brownstone residences were adjacent to each other and appeared to be of similar construction and vintage. All three measurements were made in the living room on the first floor. There was a full basement below. 5.6 Interpolation Procedure The purpose of the field measurement program was to provide estimates of the indoor levels of ground-borne vibration/noise at the residences of all 1306 telephone survey respondents. Because it was unrealistic to measure inside 1306 residences, an interpolation scheme was used to predict the vibration and noise based on measurements at a number of grid points distributed over the test area. The process was two-fold, first to estimate the exterior vibration levels at the receiver using the grid measurements, and second to adjust the exterior levels to predict the sound and vibration inside the receiver location. The exterior vibration levels were calculated using linear interpolation based on the closest neighboring grid points (some exterior residence locations were also used as grid points). Separate interpolations were done along the length of the alignment and perpendicular to the alignment. Figure 44 shows a sample interpolation calculation. The following steps formed the basis of the interpolation process: 1. The alignment was first represented as a series of straight line segments approximately 30 cm long. The latitude and longitude of each segment were calculated using aerial mapping software. For sub-surface systems, the agencies provided the necessary details to locate the centerline of the alignment. 2. Using the position of the residence, the closest point of approach (CPA) to the alignment was calculated, along with the coordinates of the CPA on the alignment. In Figure 44 for example, the residence was 76.7 m to the north of the alignment. The closest façade facing the alignment was used to identify the residence position. Residence locations were determined using aerial mapping software. 3. The closest grid points along the alignment were then determined based on the CPA point. In Figure 44, the closest grid points were G3 to the east and 70018 to the west. The position of the CPA with respect to the grid points was then calculated. In this case the CPA was closer to the western grid point and was 10.6% of the distance from 70018 to G3. (For a 50% ratio, the estimated vibration would be an average of the two nearest grid points.)
TCRP D-12 Final Report 86 4. The vibration levels at the two neighboring grid points were then adjusted to the same setback distance of the residence using a linear interpolation between the perpendicular grid points. In this case, the residence was 76.7 m from the alignment, which placed it between perpendicular grid points G16 (47.7 m) and G20 (84 m). The relative position ratio was then calculated based on the grid points. In this example the ratio was 79.9%, which meant the residence was closer to G20. The vibration levels at the neighboring grid points were then adjusted for distance based on this ratio. Once, the grid point levels were adjusted, the residence level was calculated based on the "along the alignment" interpolation ratio, determined from the previous step. As shown in Figure 44, G3 (4 m) was the closest perpendicular grid point to the alignment. To prevent unexpected interpolation results in the event that a residence was closer to the alignment than G3, an artificial grid point, D=0, was created at the center of the alignment. The D=0 point was given the same attenuation parameters as G3, preventing any singularities in the interpolation routine. (In essence, any residences that were closer to the alignment than G3 were given the same attenuation parameters as G3.) Similar artificial grid points were used at each end of the alignment, and at a suitably large perpendicular distance from the alignment. In each case, the artificial grid point was assigned the parameters of the closest real grid point. The interpolation procedure was used to estimate the exterior vibration level at each residence on a 1/3 octave band basis. Figure 44: Sample Residence Interpolation Calculation
TCRP D-12 Final Report 87 5.7 Summary of Measurement Observations The following points summarize the key observations regarding the D-12 field measurement data. ⢠The 1/3 octave band vibration spectra at most sites were greatest between 20 Hz and 80 Hz. Because most weighting curves used in national standards are effectively flat over this frequency range, the resulting overall vibration levels differ only by a constant. Consequently, the D-12 data does not have the necessary frequency range to permit a full comparison of the effectiveness of the different weighting curves at predicting annoyance. ⢠A-weighted velocity was less well-correlated to overall velocity than many of the others considered. Consequently, this metric had the potential to explain more (or less) of the community response variance. ⢠The average difference between outdoor and indoor vibration in the D-12 data was essentially zero. The standard deviation of the difference was approximately 5 dB, which reflects a wide building-to-building variation in the response to vibration. ⢠The average difference between the measured sound level and the indoor vibration level was about â5 dB for the audible frequency bands where the vibration was most severe. ⢠The correlation between indoor sound and indoor vibration was only slightly better than the correlation between indoor sound and outdoor vibration. This suggests that outdoor vibration was almost as good a predictor of indoor sound as was indoor vibration. ⢠Of the 34 residences where indoor measurements were performed, 11 had A-weighted ground-borne noise levels that exceeded 40 dB. Nine of the residences were in Toronto, one was in Boston near the Blue Line, and one was near a crossover on the Sacramento Regional Transit Gold Line. ⢠The experience from this project is that measuring noise and vibration inside of a residence is much more difficult and time consuming than measuring the vibration levels immediately outside the residence. The extra effort and time is primarily related to obtaining the necessary approvals from property owners, scheduling appointments with built in time buffers, and set up and removal of the equipment. Also, the field crew has to be larger to accommodate indoor and outdoor measurements. ⢠Assuming equal average vibration levels, there would be approximately a 10 decibel spread in measures of 24-hour exposure at the different transit systems. Because of longer and more frequent trains, vibration exposure in Toronto, Boston and New York would be 8 to 10 decibels greater than in Dallas and Sacramento. This effect is illustrated in Figure 45. The figure compares the calculated 24-hour exposure at each D-12 residence to the average train vibration at each residence. For an average train vibration level of 65 VdB, for example, the 24-hour exposure was approximately 43 VdB in Dallas and Sacramento, 50 VdB in Boston, and 52 to 54 VdB in New York and Toronto.
TCRP D-12 Final Report 88 Figure 45: Calculated Vibration Exposure at D-12 Residences based on Vibration Level Train Leq is the RMS average vibration over the 3 decibel down points. âExposureâ is the equivalent vibration level from train traffic distributed over 24-hours (analogous to Leq(24) used to evaluate community noise).
