Shock and Vibration

Volume 2016 (2016), Article ID 6031893, 21 pages

http://dx.doi.org/10.1155/2016/6031893

## Research on Aerodynamic Noise Reduction for High-Speed Trains

State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu 610031, China

Received 3 June 2016; Accepted 18 August 2016

Academic Editor: Salvatore Russo

Copyright © 2016 Yadong Zhang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

A broadband noise source model based on Lighthill’s acoustic theory was used to perform numerical simulations of the aerodynamic noise sources for a high-speed train. The near-field unsteady flow around a high-speed train was analysed based on a delayed detached-eddy simulation (DDES) using the finite volume method with high-order difference schemes. The far-field aerodynamic noise from a high-speed train was predicted using a computational fluid dynamics (CFD)/Ffowcs Williams-Hawkings (FW-H) acoustic analogy. An analysis of noise reduction methods based on the main noise sources was performed. An aerodynamic noise model for a full-scale high-speed train, including three coaches with six bogies, two inter-coach spacings, two windscreen wipers, and two pantographs, was established. Several low-noise design improvements for the high-speed train were identified, based primarily on the main noise sources; these improvements included the choice of the knuckle-downstream or knuckle-upstream pantograph orientation as well as different pantograph fairing structures, pantograph fairing installation positions, pantograph lifting configurations, inter-coach spacings, and bogie skirt boards. Based on the analysis, we designed a low-noise structure for a full-scale high-speed train with an average sound pressure level (SPL) 3.2 dB(A) lower than that of the original train. Thus, the noise reduction design goal was achieved. In addition, the accuracy of the aerodynamic noise calculation method was demonstrated via experimental wind tunnel tests.

#### 1. Introduction

With the high running speeds of high-speed trains, problems that can be neglected at low speeds become sufficient to limit improvements in train speed [1, 2]. Shen [3] notes that the dynamic environment of an ordinary train depends mainly on machinery and electricity. By contrast, the dynamic environment of a high-speed train depends mainly on aerodynamic forces; therefore, aerodynamic noise becomes the greatest limitation. Thompson et al. [4] also indicate that noise pollution has become the most critical environmental problem along high-speed railways. Aerodynamic noise rapidly increases with increasing running speeds, becoming the main noise source for high-speed trains. Indeed, the sound power of aerodynamic noise grows as the 6th power of the running speed [5]. Noise that exceeds standard requirements has become one of the main limiting factors of train speeds, restricting the sustainable development of high-speed railways. Thus, studies of the characteristics of aerodynamic noise, low-noise designs, and noise reduction can facilitate the further development of high-speed trains.

In current research on aerodynamic noise, the location and classification of the noise sources and related low-noise design principles are the main research focus. Talotte [1], Nagakura [6], and Kitagawa and Nagakura [7] report that the main aerodynamic noise sources for high-speed trains are as follows: the pantograph, the first bogie, the power head nose, the head cowcatcher, the train head, the train tail, the train windows, the train doors, the inter-coach spacings, and the bogie skirt boards. These aerodynamic noise sources can be classified into two types. The first type corresponds to noise radiated by a steady flow structure. For example, the steady vortex shedding immediately behind the pantograph can generate significant aerodynamic noise, which strongly contributes to the overall noise. Moreover, certain cavity structures on the surfaces of high-speed trains can also generate aerodynamic noise. The other type of noise source refers to noise emitted by turbulent fluctuations, which occur mostly in the turbulent boundary layer near the surface of a high-speed train or at locations where flow separations occur.

Reducing the aerodynamic noise of high-speed trains requires reducing the noise from the main noise sources. The significant current collection between the pantograph and catenary set on top of a high-speed train, because of its complex structure, strongly affects the train’s aerodynamic performance. Rough areas on the train severely disturb the airflow at high speeds, thereby generating complex flow separation and vortex shedding phenomena; the resulting powerful fluctuations in air pressure in the far field translate into aerodynamic noise. This source can be predominant in case of the high-speed train is running behind a ~4 m high noise barrier as the pantograph is higher than the noise barrier. King III [8] presented a method of using dipole sources to describe the aerodynamic noise induced by the vortex shedding of a pantograph and discovered a linear relationship between the far-field aerodynamic noise and the logarithm of the speed. Noger et al. [9] performed wind tunnel experiments on a 1 : 7 scale mock-up both with and without pantographs and identified the pantograph fairing as an important aerodynamic noise source. Sueki et al. [10] conducted full-scale and wind tunnel measurements of a PS207 pantograph (as installed on the Series E2-1000 Shinkansen) equipped with porous materials and demonstrated that a noise reduction of 1.9 dB(A) at 360 km/h could be achieved, with a significant noise reduction of 5 dB at the 1/3 octave band centre frequency at 250 Hz. The results show that the material properties of a pantograph significantly influence its aerodynamic noise. Lee and Cho [11], Ikeda et al. [12], and Kurita [13] developed an optimized pantograph structure and tested the proposed PS207 pantograph in a noise reduction experiment, in which the noise reduction effects of a low-noise panhead structure in combination with the new pantograph type were studied in wind tunnel tests. Yu et al. [14] also considered the effect of a pantograph fairing installed on the roof of a train. They found that if the fairing consisted of two sideward baffles acting as noise barriers, then a noise reduction of 3 dB could be achieved.

The bogie is one of the main aerodynamic noise sources for a high-speed train. However, because of the complex structure and disordered distribution of the flow field around a vortex, bogie aerodynamic noise is poorly understood. In numerical calculations, only minimal noise reduction can be achieved if the bogie structure is excessively simplified or if a simplified small-scale model is used. Wakabayashi et al. [15] and Kurita et al. [16] performed noise tests on a FASTECH360 S train with full skirt boards for all bogies and found that the noise from the lower part of the train was reduced by approximately 1 dB compared with a Series E2-1000 train. Based on a delayed detached-eddy simulation (DDES) and the Ffowcs Williams-Hawkings (FW-H) equation, the distribution patterns of the dipole and flow-field characteristics of a simplified 1 : 10 scale bogie model including a wheel set and bogie frame were predicted and investigated by Zhu et al. [17]. The accuracy of the numerical calculations was verified via wind tunnel experiments. Zhang et al. [18] presented a numerical method of using dipole sources to describe the aerodynamic noise induced by the trailer car bogie and discovered that the far-field aerodynamic noise of the trailer car bogie was broadband noise that has noise directivity, attenuation characteristic, and amplitude characteristic.

A low-noise structure for the CRH3 train was designed using both a nonlinear acoustic solver and the FW-H equation by Sun et al. [19]. Their results showed that the use of smooth transitions for head windows and streamline positions and the use of a streamlined design for the head cowcatcher could reduce the near-field noise by 7 dB(A) and 14 dB(A), respectively, compared with the original model. Yamazaki et al. [20] found that the inter-coach spacing is also a major noise source for high-speed trains by conducting wind tunnel experiments and field tests using a 1 : 5 scale Shinkansen train model.

Currently, research on aerodynamic noise related to high-speed trains is primarily conducted through experiments and numerical simulations. The former predominantly consist of wind tunnel experiments based on reduced-scale models and field tests based on full-scale models. However, field tests are time consuming, require high levels of manpower and resources, and are subject to numerous restrictions in terms of simulating real-world conditions. Meanwhile, wind tunnel experiments based on reduced-scale models must satisfy strict test conditions (such as conditions on the Reynolds number and the turbulence intensity of the boundary layer). Therefore, with the development of modern computers, numerical simulations have gradually become an accepted means of predicting aerodynamic noise. Compared with experiments, numerical simulations offer greater controllability and better facilitate the calculation of aerodynamic noise. Moreover, they allow noise to be predicted under different flow conditions and for different parameters values.

As an extension of the investigations discussed above, the study presented in this paper addresses the aerodynamic noise induced by full-scale high-speed trains, the characteristics of the far-field aerodynamic noise, and the study of noise reduction. To obtain high-quality results for high-speed trains through numerical simulations, a highly detailed geometry was used, and the unsteady simulations were performed at full scale. An aerodynamic noise model was established for a full-scale high-speed train, including three coaches with six bogies, two inter-coach spacings, two windscreen wipers, and two pantographs (with the first pantograph folded and the second pantograph lifted on the middle coach). The steady-state RNG turbulence model and broadband noise sources were used for a preliminary study of the aerodynamic noise sources. Then, by combining the model with an unsteady DDES and the FW-H equation, the far-field aerodynamic noise was analysed. Based on the studies described above, an accurate distribution of the aerodynamic noise sources on the surface of the high-speed train and the far-field distribution characteristics were obtained. Low-noise design principles were developed and noise reduction analyses were conducted for the main aerodynamic noise sources. To reduce noise, it is essential to consider the two main operation orientations of a pantograph, different pantograph fairing structures, different pantograph fairing installations, different choice of which pantograph is lifted in double pantograph, the presence of diaphragm plates in inter-coach spacings, and the effects of bogie skirt boards. Wind tunnel tests were performed to thoroughly verify the correctness of the numerical analysis method used in this paper.

#### 2. Numerical Fluid Analysis and Aerodynamic Noise Analysis Methods

##### 2.1. Delayed Detached-Eddy Simulation

Computational fluid dynamics (CFD) was used to calculate the induced airflow around the train, and computational aeroacoustics (CAA) was applied to predict the aerodynamic noise [21]. In typical CFD turbulence models, a detached-eddy simulation (DES) model is more universal than an unsteady Reynolds-averaged Navier-Stokes (URANS) model or a large eddy simulation (LES) model and saves significant computation time compared with direct numerical simulation (DNS) or the LES approach [22]. Therefore, dynamic fluid flow characteristics can be obtained using a DES model, whereas a URANS model with empirical parameters can capture only unsteady mean-flow structures and is therefore unable to provide the detailed unsteady flow information required for noise calculations [4]. DNS is considered to be a simple research tool that is suitable for scenarios with low Reynolds numbers [22, 23]. Although LES is much cheaper than DNS, it still requires a significant number of grid points, especially near solid boundaries, and must be run with very small time steps, which means that it is not a practical option for most industrial applications [24]. Based on the above considerations, a DES turbulence model is clearly the most suitable means of aerodynamic noise prediction for high-speed trains at present. A refinement of DES is the delayed detached-eddy simulation (DDES) approach, which was developed to avoid grid-induced separation and to preserve URANS modelling throughout the boundary layer [25]. There are two popular types of DDES model, one based on the simple* Spalart*-*Allmaras* model and one based on the SST* k*-*omega* model, of which the latter is used in this paper.

##### 2.2. Ffowcs Williams-Hawkings Acoustic Analogy

The acoustic analogy method is widely used in CAA. The theory was first presented by Lighthill [26] and was extended by Curle [27]; subsequently, Ffowcs Williams and Hawkings [28] developed the FW-H equation. This differential equation [28] is given as follows:where is air density, air pressure, sound pressure, normal direction, constant speed of sound in the undisturbed medium, time-averaged velocity in the direction, time-averaged velocity in the direction, surface velocity component normal to the surface, Lighthill stress tensor, fluid compressive stress tensor, Dirac delta function, and Heaviside function.

Based on the near-field unsteady flow data obtained from CFD calculations, the FW-H equation can be used to predict the sound generated from equivalent acoustic sources. The right-hand side of (1) includes contributions from the monopole sources, dipole sources, and quadrupole sources. When a high-speed train runs at a certain speed (up to a speed of 400 km/h, at a Mach number () of 0.33), monopole sources need not be considered because the train’s surfaces can be regarded as arbitrary rigid bodies and the pulsating volume approaches zero [19, 21]. The authors of [29] note that the intensity ratio of between quadrupole and dipole sources intensities in the flow field is proportional to . Because high-speed trains still operate at a sufficiently low speed that the noise intensity of quadrupole sources is far less than that of dipole sources, quadrupole sources can also be ignored. Therefore, the problem of the far-field aerodynamic noise induced by dipole-type sources on high-speed trains is considered in this work.

#### 3. Evaluation Criteria for Low-Noise Design

In this paper, high-speed railway noise is evaluated in terms of the A-weighted equivalent continuous sound pressure level (SPL) (). According to the definition given in the international standard ISO3095-2013 [30] for the noise testing of high-speed trains, is calculated as follows:where is a time interval, is the instantaneous A-weighted SPL, and is the reference sound pressure of 20 *μ*Pa.

The fast Fourier transform is used to transform the sound pressure at far-field evaluation points into the frequency domain, where they are revised using the A-weighted frequency-domain table [31]. Then, the inverse Fourier transform is used to transform the frequency-domain sound pressures into the time domain to obtain for the evaluation points. Finally, the values for the evaluation points are calculated using (3). Two derived quantities based on are used as indices for aerodynamic noise assessment: the max () and the average (). According to the principle of energy superposition, is expressed in the following form:where () denotes the value for the th evaluation point and is the number of noise evaluation points.

#### 4. The Aerodynamic Noise Calculation Model

##### 4.1. Computational Model

A specific type of full-scale high-speed train was selected as the research object for this study. This train is equipped with three coaches, namely, the head train, mid train, and tail train, and includes six bogies, two windscreen wipers, two inter-coach spacings, and a pantograph fairing and two pantographs on the mid train (the first pantograph is folded, and the second is lifted; see Figure 1(a)). The vehicle parameters are as follows: the model has full-scale length, width, and height dimensions of 80.89 m, 3.36 m, and 3.86 m, respectively. The head train and tail train are both 27.12 m in length, and the shape of the tail train is the same as that of the head train. The length of the mid train is 25 m. The length of the streamlined head is 5.48 m, the horizontal maximum cross-sectional area of the train is 12.16 m^{2}, and the slenderness ratio is 1.63. The original pantograph fairing design is denoted by the code Dlz1 (see Figure 1(b)).