Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 Porto, Portugal, 30 June - 2 July 2014 A. Cunha, E. Caetano, P. Ribeiro, G. Müller (eds.) ISSN: 2311-9020; ISBN: 978-972-752-165-4 Prediction model for train induced vibration and structural noise in buildings 1 H. Gjelstrup1, J. Andersen1, A Larsen1, J. Sandreid2 COWI A/S, Parallelvej 2, 2800 Kongens Lyngby, Denmark. Parallelvej 2, 2800 Kongens Lyngby, Denmark. 2 Banedanmark, Vasbygade 10, 2450 København SV, Denmark. email: [email protected], [email protected], [email protected], [email protected] ABSTRACT: A general model for prediction of ground borne vibrations and structural noise in buildings in relations to passing trains is introduced. The model is intended for estimation of vibration and structural noise levels in buildings from nearby rail lines and for assessment of human discomfort in relation to existing well established comfort boundaries. The human comfort part of the prediction model is based on transfer functions between ground, foundation and floor structures, whereas the structural noise part is based on a correlation between measurements made on the floor in the same room as the noise was recorded. The transfer functions, which are used in the model is derived from measurements of several train types and at several geographical locations, representing most of the rail systems used in Denmark. Apart from Danish locations, data from Austria is also included in deriving the model. The structural noise part of the model is based on measurements in 26 different houses, where the rooms were chosen based on minimizing external sound sources. Comparisons of predicted and measured values are presented in conjunction with theoretical expression used in the prediction model. The model show good results in predicting human comfort levels, whereas the prediction of the noise level leads to the need for a more in depth analysis of the measured data. KEY WORDS: Prediction; Train induced vibrations; Human comfort; Structural noise. 1 INTRODUCTION A general model for prediction of ground borne vibrations and structural noise in buildings in relations to passing trains is introduced. The model is intended for estimation of vibration and structural noise levels in buildings from nearby rail lines and for assessment of human comfort in relation to existing well established comfort boundaries. The model is based upon classical empirical assumptions leading to adding up frequency dependent transmission loss functions from source to receiver in order to estimate vibration levels from passing trains within a building. The model is derived from measurements of several train types and at several geographical locations, representing most of the rail systems used in Denmark. Apart from Danish locations, data from Austria is also included in deriving the model. The collection of the experimental data was carried out using two main measurements campaigns. The first part was based on finding geological damping and transmission loss functions for vibration from passing trains to houses, whereas the second part was based on expanding the transmission loss functions database from ground to the house and finding a relationship between measured vibration level on the floor and sound pressured in the same room. 2 and at some locations only vibrations at the rail line or in the building has been measured. 2.1 Part one The generic instrumentation setup consisted of accelerometers, microphones, temperature sensors and ultrasonic distance sensors for determination of train speed and length. Vibrations were measured by accelerometers at the rail, sleeper, ballast and 7.5m, 15m, 30m from the rail center. Furthermore a microphone was placed a 7.5m from the rail center and inside one family houses, where vibrations also were measured by accelerometers. Figure 1 show the geographical locations in Denmark where data has been measured. Apart from these locations, data from measurements carried out in Austria, are also included in the model. For the Kværkeby location ( ) a setup involving 72 geophones was placed in groups of 4 with ½ m spacing perpendicular to the track. SELECTED MEASUREMENT LOCATIONS AND SETUP The following section describe the measurement setup used during the two main part of the measurement campaign. The measurement setup for the two parts of the campaign was generally carried out with the same measurement setup. However some minor variation occurs due to local limitations Figure 1. Measurement locations in Denmark, from where data is included in the model. Each bullet represents several local positions. 819 Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 Hammer blows on the surface and rail as well as vibrations from passing trains was recorded in order to conduct a seismic survey and thereby understand the wave propagation in the geology (boulder clay on top of lime stone, typical Danish situation). A list of typical sensor locations is shown in Table 1 Table 1. Sensors and positions used as generic setup. was a straight line between the vibrator the 5m point and the point on the foundation. Table 2 show a list of the used sensors and their location in the measurement setup Table 2. Sensors and positions used in the second part. Sensor Position accelerometer 5m from house foundation accelerometer At the house foundation Sensor Position Sensor Position accelerometer Ground floor Thermometer Rail accelerometer At building accelerometer foundation microphone accelerometer Rail accelerometer Ground floor accelerometer Sleeper accelerometer 1. floor accelerometer Ballast Microphone Ground floor, accelerometer 3 the floor 7.5m from Ultra-sonic At sleeper center of microphone 10m down nearest from rail track accelerometer 7.5m from Ultra-sonic At sleeper center of 10m up from nearest rail track and accelerometer 3.1 Model composition The model is based upon classical empirical assumptions leading to adding up frequency dependent transmission loss functions from source to receiver in order to estimate vibration levels from passing trains within a building [1] & [2]. The acceleration level is measured in m/s2 and converted into 1/3-octave spectrum in dB with a reference of 10-6 m/s2. 1,2m over Laj Lak Lh TLg TLb TLe rail level accelerometer 20m from On the floor under the microphone MODEL FOR PREDICTING HUMAN DISCOMFORT IN RELATION TO VIBRATIONS FROM PASSING TRAINS 1.2m over accelerometer 1st floor A room with no ventilation and no windows if possible Thermometer 7.5m from center of center of nearest nearest track track and in shadow where: Laj (1) Vibration spectrum in the building (1/3-octav spectrum). 5m from building foundation A more detail description can be found in [3]. 2.2 Second part The second part of the measurement campaign was perform in 26 houses located as shown in Figure 2 j is the location. Lak Lh Vibrations source from the train (1/3-octav spectrum). Correction for the train speed (1/3-octav spectrum). Lh TLg Damping relation to vibrations transmission through the ground (1/3-octav spectrum). TLb Damping relation the vibrations transmission from ground to spectrum). TLe building foundation (1/3-octav Damping in relation to vibrations transmission through the different building levels (1/3-octav spectrum). Error on the calculation. Figure 3 show a sketch of the model, and how the different terms relates to the different position from the vibrations source (the train) to the house. Figure 2. Measurement locations for the second part of the measurement campaign. For the second part the instrumentation setup consisted of accelerometers and one microphone. The accelerometer was place at 5m, from the house foundation, on the foundation, at the ground floor and at the 1st floor if possible. The vibration source for theses measurement was a seismic vibrator which was place 5 to 10 from the 5m point and was lined up so there 820 TLe Lak Lh TLg Figure 3. Sketch of model. TLb Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 The new feature is to base the statistical calculations behind the model upon a very large set of input data, detailed analysis of train speed effects and soil/distance damping by analytical 1 or 2 layer models based upon input from geotechnical borings. The assumption that the individual 1/3 bins are normal distributed is based on a "Chi-Square Goodness of Fit Test" on the dataset of each 1/3-octave bin. Using the above mention assumption it is possible to rewrite equation (1) into the following expression of a normal distribution. N TL , TL N TL , TL N Laj , Laj N Lak , Lak N Lh , Lh N TLg , TLg b b e (2) e where: Laj Lak Lh TLg TLb TLe Figure 4. Number of trains, by type and measurement location. The two locations is Bedstedvej 16 and Kongstedvej 15 located in Kværkeby, which was marked by in Figure 1. 2 2 2 Laj L2ak 2L TL TL TL g b e h It is assumed that the correction for the train speed and the damping in the ground is deterministic functions, why: 2 0 . This leads to a simplify expression: 2L 0 , TL g h N Laj , Laj N Lak , Lak Corr N TL , TL N TL , TL b b Where the data used for determining e e N Lak , Lak (3) Corr is corrected for the train speed and then for the damping in the ground. The model works by choosing a set of six parameters which defined the vibration case which is to be studied. The six parameters are: Rail track type. Rail track level in relation to the surroundings. Train type. House type. Floor level. Eigenfrequency of the floor (4 frequency intervals). For a more detail description of the human comfort model, please see [3]. 4 4.1 MEASURMENT RESULTS Figure 5. Two selected locations in Kværkeby: Bedstedvej 16, = Kongstedvej 15. = Figure 5 show the two locations and how the alignment of rail track is between the two locations. In order to make a comparison between the two locations, all trains measured at Kongstedvej 15 was also measured at Bedstedvej. The difference in the number of measured trains, which is seen in Figure 4, is due to the fact that the measurements started at Bedstedvej 16 before Kongstedvej 15. Figure 6 show the distribution of the train speeds of the measured trains at Bedstedvej 16. This particular part of the rail track is has a max speed of 180 km/t which the measured results also indicates. Results from passing trains In the following part selected result from part one of the measuring campaign is shown. Figure 4 show the distributions of the measured trains for two locations which is about 2 km apart. It is seen that there at two main types of trains namely IC3 and ER. These two train types are more or the same except IC3 is powered by fossil fuel and ER is electrified. In the following the results from the IC3 trains is shown in greater detail. Figure 6. Train speeds measured at Bedstedvej 16, by train type. Figure 7 (top) shows the measured source strength 7.5 m from the track center at Bedstedvej 16. A dominant frequency is identified in the range 60-80 Hz. Figure 7 (bottom) shows 821 Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 houses are significantly different from the other house which was measured. Transmission loss the measured source strength 7.5 m from the track center at Kongstedvej 15. There is again a dominant frequency range 60-80 Hz. It is also seen that the levels are approx. 10 dB lower than at Bedstedvej 16. Hz Figure 8. Transmission loss from 5m to foundation. Bold line is showing the mean curve. Transmission losses This section presents only data from part two of the measurement campaign. In part two of the measurement campaign, where the vibrations source was a seismic vibrator, measurements in 26 houses was preformed, in order to identify the transmission loss from ground to house foundation, from house foundation to the ground floor and from the house foundation to the 1st floor. Figure 8 show the transmission loss from 5m from the house and to the house foundation. The transmission loss is defined as measurements at the house foundation divided by the measurements at the 5m point, within a 1/3 octave spectrum. It seen there is a quit large difference in the results between the individual measurements. It seems like there are two sets of measurements which fall outside the grouping of the other. The reason to this difference has not been found, but it is believed that the geology or the foundation of for these two 822 Transmission loss 4.2.1 Results from seismic vibrator Hz Hz Transmission loss 4.2 Transmission loss Train speed was measured for each train passage at both Bedstedvej 16 and Kongstedvej 15 and significant difference was found in the in the speed of the same train at both sites. Based on this observation, it is assumed that the differences seen between the two sets of curves shown above may come from a difference in reaction forces on the track and/or differences in ground conditions. The biggest difference is seen in the frequency bands between 1.25 and approx. 90 Hz, where the level measurements are approx. 10 dB higher for Bedstedvej 16 than for Kongstedvej 15. The variations in the individual curves in the top and bottom of the figure are due the different speed of the measured trains. Transmission loss Figure 7. Source strengths from IC3 trains at Bedstedvej 16(top) and Kongstedvej 15(bottom), measured 7.5 m from the centre of the track. Figure 9 and Figure 10 show obtained transmission loss from the house foundation and to the ground floor and the 1st. floor respectively. The values presented in the figures are amplifications factors to the individual 1/3 octaves in linear scale. The results are sorted into the four frequency intervals namely, 0-20Hz, 20-40Hz, 40-65Hz and 65-110Hz. The reason for sorting the transmission losses into theses interval is bases on already obtained data which have been analyses. This analyse showed that the transmission losses tended to fall in the four frequency intervals as noted above. Furthermore it shows that the amplification was link to the eigenfrequency of the floor. In comparison to the transmission loss from the ground to the foundation, then the result found for the transmission loss from the foundation to the different floors are nicely group in three frequency intervals, namely 20-40Hz, 40-65Hz and 65110Hz. From the Figure 9 and Figure 10 it's seen that none of the measured transmission losses falls within the 0-20Hz interval. Hz Hz Figure 9. Transmission loss from foundation to ground floor, sorted after amplification in 4 pre-set frequency intervals. Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 Hz Hz LAeq/Afloorarea SoundPress(A)/floor 25,00 Transmission loss Transmission loss Floor; Concrete 20,00 15,00 10,00 5,00 0,00 Hz Hz For the ground floor it seems that most of the measured transmission losses falls within the 65-110Hz interval, see Figure 9, whereas for the 1st. floor there seems that the measured transmission losses is mainly distributed in the 4065Hz interval, whereas the rest is distributed between 2040Hz and 65-110Hz. Looking at the distribution of transmission losses within these intervals, it is seen that even though there is some deviation from the mean, then there is a definite trend within the groups shown in Figure 9 and Figure 10. Structural noise Structural noise was also measured during the measuring of the above shown transmission losses. As mentioned above the instrumentation setup was a microphone placed 1.2m above an accelerometer which was placed on the floor. The data from the both sensors was analyses and converted into 1/3 octaves running from 10Hz to 160Hz, in dB, which is the range defining the structural sound in Denmark, Erro! A origem da referência não foi encontrada.. The data recorded by the microphone have been A-weighted and the data from the accelerometer have been KB-weighted. Figure 11 and Figure 12 shows the relationship between the structural noise and the measured vibration level on the floor, which was found for a concrete floor and at wooded floor respectively. The plots show the weighted total band power which have been normalized with the floor area of the room in which the measurements was obtained. The total band power(LAeq and Law) used for producing Figure 11 and Figure 12 is calculated by: 20Log10x / 1e 6 n where x is given by 10,0 15,0 20,0 25,0 30,0 Figure 11. Relation between acceleration and sound pressure measured on a concrete floor Figure 10. Transmission loss from foundation to 1st. floor, sorted after amplification in 4 pre-set frequency intervals. 4.2.2 5,0 Acc band power(KB)/floor Law/Afloor area Making a linear fit to the data found on a concrete floor, which is presented in Figure 11, result in the expression for predicting the A-weighted sound pressure as a function of the KB-weighted acceleration and the rooms floor area, see equation (4). LAeq 0.8136 Law 0.7666 Afloor (4) Floor; wood 16,00 LAeq/Afloorarea SoundPress(A)/floor Transmission loss Transmission loss 0,0 14,00 12,00 10,00 8,00 6,00 4,00 2,00 0,00 0,0 5,0 10,0 15,0 20,0 25,0 Acc band power(KB)/floor Law/Afloor area Figure 12. Relation between acceleration and sound pressure measured on a floor made out of wood. The distribution and the of number data points for the wood floor makes it difficult to say anything certain about the relationship, but under the assumption that it following the same trend as for the concrete floor the following has been found. Making a linear fit to the data found on a wooded floor, which is presented in Figure 12, result in the following expression for predicting the A-weighted sound pressure as a function of the KB- weighted acceleration and the rooms floor area, see equation (5). . LAeq 0.5847 Law 0.4061 Afloor (5) Octave i2 , were i is the Octave bin i 1 running from 10Hz to 160Hz and the octaves is either Aweighted or KB-weighted. Looking at the measurement in a combined plot yields the result shown in Figure 13 and equation (6). 823 Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 5.3 Floor; Concrete and wood LAeq/Afloor area SoundPress(A)/floor 25,00 For the presented case we have a good agreement between the measured and the predicted human comfort level. The differences for the two floors were found to be: 20,00 15,00 • • 10,00 5,00 0,00 0,0 5,0 10,0 15,0 20,0 25,0 30,0 Acc band power(KB)/floor area Law/Afloor Figure 13. Relation between acceleration and sound pressure measured on floors made out of wood and concrete. LAeq 0.7707 Law 0.3709 Afloor (6) More measurements in room with floors made of wood needs to be performed in order to say if there are different expressions as shown in equation (4) and (5) for different floor types or if a combined expression can be made as shown in equation (6). Ground floor: 1st floor: 5.4 1.2 dB -1.1 dB Structural noise level bases on measurements. Using a passing train as the vibration source the following noise and floor vibrations was measured in the same room. The room has a concrete floor with an area of 5.95m2. The room was empty but has a door leading outside with a large window, which was cover with Rockwool to minimize airborne sound. • • Floor vibration level: Noise level: 5.5 58.29 dB(KB) 34.27 dB(A) Structural noise level bases on model predictions Using equation (4) and the floor vibration level mention above result in the following prediction of the structural noise. • 5 Comparison of measured and predicted human comfort Predicted noise level: 42.9 dB(A) MODEL PREDICTIONS In the following a presentation of measured and model predicted Human comfort levels and structural noise is given. 5.1 Human comfort level bases on measurements. Looking at measurements performed at a house located about 30 meters from the rail track; the following human comfort level is calculated from measured data obtained in the ground floor and at the 1st floor. The measurements were performed during the fall of 2012 and the values are given as a 95% confidence interval. • • Ground floor: 1st floor: 72.9 dB(KB) 76.5 dB(KB) 5.6 Comparison of measured and predicted noise level Looking at the measured and predicted noise level, it's found that there is a relative large difference of about 9 dB in over estimating the measured value. Looking at the measurements which was used in deriving the model in compassion to the measurements used in prediction the noise level shows a discrepancy which seems to be the reason to the offset in the predicted value, see Figure 14 and Figure 15. 3,50E-04 3,00E-04 2,50E-04 The house is a single family house and the first mode for both floors falls within the interval of 20 Hz to 40 Hz. The data used to calculate the human comfort level originates from passing train of the type MF, also called IC3, and the rail track itself was situated about one meter above the surrounding landscape. 5.2 Human comfort level bases on model predictions Using the model and inputting parameters which corresponds to the case described above yields the following results, also presented in a 95% confidence interval: • • 824 Ground floor: 1st floor: 71.7 dB(KB) 77.6 dB(KB) 2,00E-04 Accelerometer 1,50E-04 Microphone 1,00E-04 5,00E-05 0,00E+00 10 100 Hz Figure 14. Measurements used in deriving the prediction model. Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014 5,00E-03 4,50E-03 4,00E-03 3,50E-03 3,00E-03 2,50E-03 Accelerometer 2,00E-03 Microphone 1,50E-03 1,00E-03 5,00E-04 0,00E+00 10 100 Hz Figure 15. Measurements used in predicting the noise level. A deeper analysis of is discrepancy is still ongoing. 6 CONCLUSION Using a general empirical model for vibration transferring from a source to a location, the human comfort level is calculated in a 95% confidence interval. Furthermore a linear relationship between the structural noise and the vibration on the floor has been presented. A comparison between the values obtained from a calculated 95% confidence interval and the measured values shows a good agreement between the model and measured values. The difference is within ±1.2 dB for the case presented in this paper. A comparison of measured noise level and predicted noise levels has shown a relative large discrepancy. A deeper analysis of why discrepancy is found is still ongoing. ACKNOWLEDGMENTS I would acknowledge Rail Net Denmark for giver us this opportunity for make this model and letting us perform this extensive measurement campaign. REFERENCES [1] [2] [3] [4] VIBRA-1-2-3: A software package for ground borne vibration and noise prediction , Dr. A. Ziegler / Dipl. Ing. ETH / M.Sc. UCB / Beratende Ingenieure USIC, Asylstrasse 41, CH-8032 Zürich C. Madshus, B. Bessason and L. Harvik, Prediction model for low frequency vibration from high speed railways on soft ground, Journal of Sound and Vibration (1996) 193(1), 195-203. H. Gjelstrup, A. Larsen, J. Andersen, H.B. Kock and J. Sandreid, New approach to calculated human discomfort affected by vibrations from passing trains. Proceedings from IWRN11 Information from the Danish Environmental Protection Agency No. 9/1997 825
© Copyright 2024