TRAIN INDUCED SEISMIC NOISE OF ACCELERATING AND DECELERATING TRAIN SETS

TRAIN INDUCED SEISMIC NOISE OF ACCELERATING AND DECELERATING TRAIN SETS ABSTRACT: M. Çetin 1, A. Tongut 2, S.Ü. Dikmen 3 and Ali Pınar 4 1 Civil Eng., Dept. of Earthquake Engineering, KOERI, Bogazici University, Istanbul 2 Civil Eng., Dept. of Earthquake Engineering, KOERI, Bogazici University, Istanbul 3 Assoc. Prof. Dr., Dept. of Earthquake Engineering, KOERI, Bogazici University, Istanbul 4 Prof. Dr., Dept. of Earthquake Engineering, KOERI, Bogazici University, Istanbul Email: mahir.cetin@boun.edu.tr The studies reported in the literature on train induced vibrations, were mostly performed by locating the instruments in the transversal direction to the railway system. In this study, a horizontal array parallel to the alignment of the railroad was used to record the train induced vibrations. The test site was in close proximity to the Atakoy subway station in Istanbul where subway trains travel above ground. The particular location of the array provided an opportunity to monitor the train induced vibrations of both accelerating and decelerating train sets. The data acquired was processed using frequency domain techniques using in-house computer codes developed in MATLAB. Analyses revealed that the train induced seismic noise is in the range of 8 Hz and above. Moreover, the frequency content of the vibrations is velocity dependent. KEYWORDS: Train induced vibration, seismic noise, frequency content 1. INTRODUCTION Train induced vibrations (TISN) is a widely researched topic by the people working on ground vibrations, namely man-made vibrations. Though TISN does not pose a threat to the integrity of structures, it often is a cause of discomfort and functionality of sensitive equipment. For instance, some hospital and laboratory equipment are sensitive to vibration and can be adversely affected when subject to even some small vibration. Thus, either the whole structure should be isolated or the equipment of interest. In that respect, TISN is one of the challenging vibration modes for the aforementioned type of structures besides earthquake or wind induced vibration. So, in order to protect structures or equipment from TISN, the frequency content of it in the region of interest must be determined. In the last several decades, studies have more focused scientists' attention on mitigation of waves due to railway source both theoretical and experimental analyses. Due to some of limitations associated with numerical analysis, [2, 3, and 4], performing in-situ experimental analysis of railways have become more popular. Field measurements have used as the experimental validation of a numerical model for the prediction of train induced vibrations in studies by researches [1, 5, 7]. However, many train induced seismic noise studies to date were concentrated on the transversal propagation of trains at constant speed of seismic field generated by railways, particularly on the high speed modern railways. These fields have been shown to be generated by different mechanisms at different frequency bands. Connolly et al. (2014), found that in contrast with the jointly agreed theory, the horizontal vibrations should also be taken account besides vertical components. The common conclusion of author s was that the ground vibration generated by trains is relatively connected to the ground and 1

train type and property. In basic terms, the relationship between train speed and soil wave propagation velocity is the parameter that determines the magnitude of the vibrations resulting from passing trains [3]. To the authors knowledge, there is limited published literature related to the 2-dimensional, wave propagation due to railway as well as seismic noise of accelerating and decelerating train sets. This paper presents and discusses the frequency content of accelerating and decelerating train sets utilizing measurements made on the railway at Atakoy region in Istanbul. 2. THE STUDY AREA AND INSTRUMENTATION The study was done at Atakoy, Istanbul whose vicinity is occupied mostly by residential buildings (Figure 1). Hence, the streets within the vicinity have low-level vehicular traffic. There is a highway, D100 passing by the region and a subway line just next to the highway. Westbound, D-100 goes towards the western borders of Turkey. Eastbound, it goes towards the city center and the Ancient Peninsula. The road has been built in the early 50 s and was expanded to an access controlled highway in the early 70 s during the construction of the first Bosporus crossing. The subway line operates from 6:00 am to midnight with trains running every 6 to 10 minutes in either direction. On the other hand, there is a subway station about 25 m from the first sensor of the array in the NNW direction that all the trains stop. Due to this fact, the trains while passing by the array are either accelerating or decelerating. Figure 1. Location of the highway and railway with respect to sensor array As mentioned earlier the data acquisition was performed by one array, namely horizontal, which is parallel to the subway line. The array was formed with 4 triaxial force balance accelerometers (Figure 1). The accelerometers were GURALP make type CMG5T. One sensor (#1) was placed at the edge of the highway, the remaining three sensors were distributed along the subway line with distances of 10 m, 80 m and 285 m from sensor # 1 respectively. The recordings are continuously made at a rate of 100 Hz. Figure 2 schematically demonstrates location of the array and dimensions of the test set-up. The data acquisition for this study was made during noon hours and it spanned 2 hours 10 minutes approximately. During the acquisition time, 44 train vibrations are recorded half of which are for trains going to the airport and the other half for trains going to Yenikapi. As mentioned before, since the train station is so close to the array, trains going to airport decelerate to stop at the station while trains going to Yenikapi accelerate as they move away from the station. Furthermore, the proximity 2

of highway and subway provided us with the opportunity to record seismic noises induced by metro buses and other vehicles passing just next to the subway. 3. ANALYSIS OF DATA Figure 2. Schematic picture of the horizontal array As mentioned above, one of the sensors, namely #1 was placed between the highway and subway, meaning it recorded both seismic noise induced by metro buses and trains. Thus, the data acquired by this sensor is the noisiest data among the data acquired by the other sensors which are deploying along the subway line. The speed of trains going to airport maximum when passing by the sensor # 4 and minimum while passing by sensor #1. For the trains going to Yenikapı, the situation is vice versa. Each sensor recorded three component of the seismic noise namely, east-west, north-south and vertical components. Figure 3 demonstrates the NS component of data acquired by sensors. Figure 3. The East-West Component Time series of Seismic Noise. In the Figure 3, each major spike corresponds to a pass by a train either going to airport or going to Yenikapı. The remaining majority part of the record is noise generated by the traffic on the D100 highway and ambient noise. The data was baseline corrected linearly with 794 break points since the sensors are not fixed at the ground which causes the shaking of sensors during acquisition meaning the centerline of the record is not straight but fluctuates with a long period instead. 3

In order to determine the frequency content of accelerating and decelerating trains, the data acquired is separated to get data for trains both going to airport and Yenikapı. After the separation, 22 data for trains going to airport and 22 data for trains going to Yenikapı is determined. For all sensors, the Fourier Amplitude Spectrum (FAS) of all passing trains and for all three components are calculated. Figure 5 shows the mean of FAS calculated. One can infer from Figure 5 that the frequency content of the train induced vibration is 8 Hz and above. Since the sampling rate is 100 Hz, the maximum frequency that we can see is 50 Hz. Thus, there may be higher frequencies beyond 50 Hz generated by trains. On the other hand, Figure 5 also contain the frequency content of vibration induced by metro buses. To examine the frequency content in more detail, the spectrograms of records are calculated (Figure 6). Note that from Figure 6, the frequency content of vibration induced by metro buses is between 10 Hz to 35 Hz at the region of sensor #1 while it is 10 Hz to 15 Hz at the other sensors. As mentioned before, sensor #1 was replaced close to highway so it is expected for sensor #1 to record the noise generated by metro buses. Figure 4. Speed Curves of Train Sets. (Dotted red line is speed curve of trains accelerating to Yenikapı and dotted blue line is trains decelerating to airport) The cross-correlations of data acquired are calculated to determine time delay between sensors deployed for both trains accelerating and decelerating. The speed of trains are calculated with respect to time by utilizing the time delays calculated (Figure.4). In order to determine the frequency content of train sets with respect to their speed, both the Figure 4 and Figure 5 are interpreted simultaneously. The frequency content of vertical component is dominated between 30 Hz and 38 Hz while trains passing by sensor #1 to Yenikapı meaning the trains accelerate (Figure 5 LHS). However, when the trains pass by sensor #2 and others, the frequency content of vertical component is both shifted to 40 Hz-50Hz and its amplitude decreases. This means that as trains accelerate, their vertical motion increase with lower amplitudes. On the other hand, when the decelerating trains pass by sensor #1, they generate motion that the frequency content and amplitudes are mostly the same as the motion generated by trains accelerating Since their speeds are the same as each other as passing by the sensor #1 (Figure 4). Also the amplitude of vertical component motion decrease as trains decelerates and the dominated frequency range shifted to lower values (Figure 5 LHS). 4

4. Uluslararası Deprem Mühendisliği ve Sismoloji Konferansı Figure 5. Fourier Amplitude Spectrum of vibration induced by trains going to Yenikapı (LHS) and Airport (RHS). (In each plot, the thick full black line corresponds to the vertical component of the vibration, the dotted blue line to east-west horizontal component and the dotted red line to north-south horizontal component. From top to down, sensor 1 to sensor 4). 5

Figure 6. Spectrogram of the North-South Component of Time series In comparison with vertical component, the horizontal components namely, east-west and north-south components are more irregular in terms of the frequency content change and amplitude with respect to speed of train sets. The reason is being that the horizontal components motion generated by trains depend mostly on the zigzags on the rails and the topography on which the railway lays. Not only do these two random reasons cause a speed independent frequency content, but also they result in speed independent amplitudes. In general, the frequency content of the traffic induced seismic noise is dominated by two frequency bands; the 8-15 and 30-50 Hz ranges corresponding to two distinct sources. These sources are the vehicles and metrobuses traveling along the three shoulders producing seismic energy around 10 Hz, and the train sets traveling along the nearest shoulder to stations, generating waveforms within frequency range of 30-50 Hz. A characteristic feature of these two frequency bands is that the waves associated with the cars and the trains produce mostly Rayleigh waves. Such an inference comes from by the large Fourier amplitudes of the vertical component shown in Figure 5. These results obtained from the waveforms acquired are in good agreement with previous findings suggesting that when an elastic half-space with a vertically oscillating force source on the surface is considered, the waves will be propagating radially from the source. Of the total input energy, 67 % radiates as Rayleigh (R) waves, 26% as shear (S) waves, and 7% as pressure (P) waves (Dikmen et al., 2016; Hunt and Hussein 2007; Miller and Pursey 1955). As mentioned earlier, the frequency content of train induced motion is dominated between 30-50 Hz. The data out of this frequency range is mostly composed of the motion from vehicles and metrobuses passing on highway. One can note that from Figure 7, after bandpass filtering between 0.05-30 Hz of the raw data acquired, the majority of data remained is composed of only train induced motion Figure 7 (RHS). 6

Figure 7. Comparison of filtered (RHS) and raw data (LHS) recorded from passing train sets. 4. CONCLUSIONS A study has been conducted to determine the frequency content of accelerating and decelerating train sets. Four accelerometer sensors are deployed parallel to the subway. The sensors recorded both the noise induced by traffic passing on D100 highway and train sets on subway parallel to the highway. Frequency domain analysis was performed. Analyses revealed that the frequency content of train induced vibration is in the range of 8 and above while vehicles and metro buses generate motion with 8-20 Hz. Furthermore, the dominated frequency range of vertical component of train induced vibration is shifted to higher values and their amplitudes decreases as trains accelerates. However, the horizontal components namely, east-west and north-south seem to be speed independent in terms of the frequency content and amplitudes. REFERENCES [1] Lombaert, G., & Degrande, G. (2001). Experimental validation of a numerical prediction model for free field traffic induced vibrations by in situ experiments. Soil Dynamics and Earthquake Engineering, 21(6), 485-497. 7

[2] Degrande, G., & Schillemans, L. (2001). Free field vibrations during the passage of a Thalys high-speed train at variable speed. Journal of Sound and Vibration, 247(1), 131-144. [3] Galvín, P., & Domínguez, J. (2009). Experimental and numerical analyses of vibrations induced by highspeed trains on the Córdoba Málaga line. Soil Dynamics and Earthquake Engineering, 29(4), 641-657. [4] Kogut, J., Degrande, G., & Haegeman, W. (2003, May). Free field vibrations due to the passage of an IC train and a Thalys HST on the high speed track L2 Brussels-Köln. In 6th National Congress on Theoretical and Applied Mechanics. [5] Lombaert, G., Degrande, G., Kogut, J., & François, S. (2006). The experimental validation of a numerical model for the prediction of railway induced vibrations. Journal of Sound and Vibration, 297(3), 512-535. [6] Krylov, V. V. (1995). Generation of ground vibrations by superfast trains. Applied Acoustics, 44(2), 149-164. [7] Gupta, S., Liu, W. F., Degrande, G., Lombaert, G., & Liu, W. N. (2008). Prediction of vibrations induced by underground railway traffic in Beijing. Journal of Sound and Vibration, 310(3), 608-630. [8] Dikmen, S. U. (2016). Response of Marmaray Submerged Tunnel during 2014 Northern Aegean Earthquake (Mw= 6.9). Soil Dynamics and Earthquake Engineering, 90, 15-31. [9] Connolly, D. P., Kouroussis, G., Woodward, P. K., Costa, P. A., Verlinden, O., & Forde, M. C. (2014). Field testing and analysis of high speed rail vibrations. Soil Dynamics and Earthquake Engineering, 67, 102-118. [10] Sheng, X., Jones, C. J. C., & Petyt, M. (1999). Ground vibration generated by a load moving along a railway track. Journal of sound and vibration, 228(1), 129-156. [11] Thompson, D. (2008). Railway noise and vibration: mechanisms, modelling and means of control. Elsevier. 8