Generation of a mountain-valley wind in an atmospheric boundary wind tunnel: a gust-wind generator study

Abstract

Appropriate modeling of an experimental technology is necessary in order to estimate the aerodynamic characteristic of railway trains and infrastructure (e.g., bridges). Simulation of the earth’s wind characteristics of nature is a well-established practice by using an atmospheric boundary wind tunnel. However, in the mountainous area, the wind characteristics are strikingly different from those of the plain area, the amplitude variation of wind is related to complex terrain. Compared with atmospheric boundary layer winds, which are customarily treated as stationary, winds associated with gust winds originating from mountain areas exhibit rapid changes during a short period. A lack of available field test data and testing techniques has hindered such knowledge of the effect of mountain wind on railway-related applications. To simulate the characteristics of gust winds and prepare for follow-up studies of the impact on the railway-related structures, a gust wind generator was developed in an atmospheric boundary wind tunnel — the CSU wind tunnel. Further, the performance of the gust-wind generator was studied and analyzed under the condition of the combined operation between a gust-wind generator and a wind tunnel.

1 Introduction

Over the past 10 years, the volume of railway operations increased by more than 52,000 km and the high-speed railways have more than tripled, ranking first in the world. A high-speed expansion of railway construction in and around cities has made this development trend even more evident in recent years. By the end of 2021, the mileage of China’s railways has breaking 150,000 km and high-speed railways reaching 40,000 km (Zou et al. 2022a). Geography is a vital environmental factor that must be considered for the intensive exploitation of railway transport, and the developing trend of railway networks puts a high demand on design and construction. Especially the complicated topographic features of mountainous areas pose multiple threats to the railway, and one of the significant effects is from the wind environment. In addition, the expanding railway network is a consequence of the need to pursue speed and convenience; therefore, train shuttles through various terrain scenes have become widespread. The safety of the vehicle and wind-sensitive infrastructures (e.g., bridges, towers) need to pay more attention to the wind environment of mountain-valley areas. However, the wind characteristics of the mountainous area are very different from those of the urban, countryside and seaside (Xing et al. 2021).

To study the wind characteristics in mountainous areas, field measurements, wind tunnel tests and numerical simulations are the three main methods frequently used (Li et al. 2017, b; Zhang et al. 2020, c; Jing et al. 2020). There is no doubt that field measurements are a reliable method for providing first-hand information about wind characteristics of the mountain area, however, the high cost and limited points lead the field measurements to be prohibitive. Hence, the other two approaches are more commonly seen in the research on wind characteristics in mountain terrains. Moreover, numerical simulation is a good idea for obtaining wind field information, and numerous researchers have contributed to the understanding of the wind characteristics of mountains areas (Han et al. 2018; Ren et al. 2018; Huang et al. 2018; Hu et al. 2018). However, the accuracy of the numerical algorithm and the requirement of computing resources for high-precision computation limited the progress of wind characteristics in mountain district studies.

For the simulation of the atmospheric boundary layer (ABL), a wind tunnel is a well-established practice. Several investigations have been conducted to figure out the experimental techniques to reproduce the wind field identical to that in mountain areas. Li et al. (2017a) studied the wind characteristics in three different types of upstream terrains and found that wind speeds along the bridge deck are mainly affected by the local topography. Zhang et al. (2020, b) investigated the wind characteristics in complex terrain using wind tunnel tests. The sheltering effect of the mountain was discussed since the wind fields are uneven and very different from that of flat terrain. Li et al. (2010) performed a wind tunnel experiment to investigate wind characteristics in the mountainous valley terrain. The velocity speed-up effect was observed by measuring the simulated atmospheric neutral boundary layer on the valley models. Song et al. (2020) found that shielding and channeling effects contribute to wind deceleration and acceleration respectively, and the speed-up of the valley and mountaintop is obvious. Recently, the effects of mountains terrain have been a wide concern, especially the boundary layer developed on the surface of terrain and wind distribution at bridge sites (Hu et al. 2015, 2018; Mattuella et al. 2016; Zou et al., 2022; Wang et. al., 2022).

Physical modeling for railway aerodynamics is usually classified into two similarity requirements: (1) reproducing the main properties of natural winds and (2) criteria for similarity of wind effects (He and Zou 2021a, b). The design of the wind tunnel tests and study of railway aerodynamics should meet both essential criteria as much as possible. First, the inspection analysis of simulated wind fields in the boundary-layer wind tunnel applied to the equations expressing basic conservation laws applicable to the lower atmosphere yields a significant set of criteria for simulating natural wind in the ABL. Then, various types of wedge interference generate different ABL wind profiles in the wind. Furthermore, spire, fence, and floor roughness are common fittings for simulating the wind field of local terrains. However, many previous studies showed that these experimental techniques have good applicability for sustained wind speeds, but some problems still exist in simulating the wind field of mountain areas in a wind tunnel. For example, the significant feature is the strong gust of wind compared to the other terrain (Jiang et al. 2021), as shown in Fig. 1. Gust winds in mountainous areas are transient winds with the typical non-stationary characteristics of varying wind speeds and directions and strong suddenness (Li et al. 2017a; Zhang et al. 2020a; Ren et al. 2021).

Fig. 1

Typical time history of wind speed in mountain areas (Li et al. 2022)

In summary, conventional wind tunnel approaches can obtain the wind field information about mountain areas by using terrain models and employing a combination of passive devices (e.g., wedge, floor roughness) to generate boundary layers of the same scale as the geometric scaling of structural models placed in them. However, the study of gusts in mountainous areas is still in the primary stage and the available information regarding the aerodynamics of railway vehicles and infrastructures subjected to gust winds in mountainous areas is much lacking. Significantly, the limitations of wind tunnel experimental techniques and the dilemma of scale ratio of terrain and structures lead to restrain the further study of railway aerodynamics in mountainous areas.

To address the issues, some solutions have been developed to reproduce gust wind and related effects on the structures in the laboratory using active devices over the years. Some new wind tunnels have emerged with added features, such as passive or active wind farm systems and simulators with vortical flows (Kobayashi et al. 1994; Cao et al. 2002). The new experimental technique has enabled many efforts and contributions toward understanding aerodynamics, fluid-structure interaction and relevant mechanisms. Hence, practical and rational experimental simulation technology is necessary, and the design of a new device to produce high resolution and high stabilization of the wind field has become the key technology used in wind tunnels.

Gust encounters are among the most critical load cases for railway aerodynamics and threaten running safety seriously. For example, a major accident occurred in India in 1981 when a train fell off a bridge after being hit by a strong gust of wind, causing more than 800 casualties among the passengers (Montenegro et al. 2022). In a more recent one in 2006 in Japan, a train hit by a lateral gust overturned at a low running speed (25 km/h) (Tamura 2009). In China, more than 100 vehicles flipped on the Lanzhou-Xianjiang railway line, where the wind gusts often reach 40–50 m/s (Zhang et al. 2013; Guo et al. 2015). All these wind-induced accidents indicate that the gust winds may pose a significant threat to the train’s safety and should be studied in detail. Therefore, considering that gust winds conditions are significantly different from traditional boundary-layer flows, a wind tunnel at Central South University (CSU) equipped with a gust-wind generation mechanism has been developed as part of a research program on wind engineering and railway aerodynamics.

2 CSU wind tunnel

The CSU wind tunnel is one of the experimental platforms in the National Engineering Research Center for Construction Technology of High-Speed Railway at Changsha and is highly dedicated to civil engineering and railway-related aerodynamics model testing. The CSU wind tunnel is a largeclose-circuit wind tunnel, as illustrated in Fig. 2.

Fig. 2

Wind tunnel at Central South University (He and Zou 2021a)

The CSU wind tunnel is currently being designed to accommodate 2 test sections (12 m × 3.5 m and 3 m × 3 m), which are connected to the contraction, settling, and diffuser sections. Two closed test sections (the two “yellow” areas) are available (Fig. 2). The good flow quality is adequate for wind engineering and industrial aerodynamic testing (the turbulence level is less than 0.3%). The wind tunnel is equipped with six large fans (3 × 2) that supply the power for the airflow movement. The airflow is streamlined by passing through turning vanes and contraction sections to the relatively large test section (12 × 3.5 m), named the low-speed section, which is controlled in velocity (from 0 m/s up to 20 m/s). On the other side, a smaller square section with dimensions of 3 × 3 m is named the high-speed section (due to the maximum velocity of 94 m/s). Furthermore, except for the wide scope of wind speed and the relatively large size, another advantage of the CSU wind tunnel is that it is convenient to adjust and maintain.

3 Design of gust-wind generating system

Compared with boundary-layer winds generally regarded as stationary, gust winds exhibit distinct non-stationarity, that is, rapid wind speed changes during a short time interval. The goal of the experimental facility was to generate gust winds in the wind tunnel test sections with the capability of large magnitude velocity changes within the shortest possible time. However, with the traditional setup of conventional wind tunnels, reproducing wind fields, testing loads and more in a general aerodynamics-related mountain-valley area inside an atmospheric boundary wind tunnel is always a challenging problem, especially when the focus of the testing activity is trying to reproduce gust wind flow and induced-effects over the structural models being analyzed. Therefore, the stationary wind model typically used in boundary-layer winds may not be valid for gust winds described in terms of time-varying parameters (Kwon and Kareem 2009). A solution has been easily found by designing, realizing and installing a gust-wind generator right at the front end of the wind tunnel test section (Kobayashi and Hatanaka 1992; Butler and Kareem 2007). As shown in Fig. 3, the proposed gust generator device is composed of 10 vanes, characterized by a symmetric airfoil. Flapping of the 10 vanes generates redirected airflow, and if the redirected airflow is controlled effectively, it is possible to generate either a gust of wind or air turbulence.

Fig. 3

Design sketch of gust wind generator

The square frame is made of rectangular steel and extruded profiles and is designed to be fixed to the wind tunnel with a cross-section 450 mm deep (i.e., along the direction where the airflow is moving downstream). The frame is joined using bolts, T-slot nuts, brackets, and specially designed attachment plates and supports the gust vanes and the actuators.

Each gust vane is made of light steel. This material allows for a lightweight design and offers good machinability. The surface is impregnated with epoxy resin to increase its toughness and smoothness. During the gust-wind generator working, the gust vanes need to bear strong inertial and aerodynamic forces. Hence, the vanes must be ensured that no bend or twist excessively in the course of the experiments to prevent detrimentally affecting the flow quality. Therefore, the airfoil-shaped vanes are hollow with a certain thickness so that an aluminum alloy tube can be embedded to provide the required strength, stiffness, and stabilization. The vanes are 2.88 m long to cover the full width of the wind tunnel section and allow connection with the actuation system.

The actuation system and vanes are connected by a special coupling joint named “steering linkage system”, which ensures high torsional rigidity. Moreover, the usage of an independent PID regulator, to avoid possible inconsistency of inertias, frictions and shape between the 10 vanes, guarantees the best mechanical rhythm to move the 10 gust vanes. Finally, a “steering linkage system” mechanism transforms the power imparted by the actuator into a vertical translating motion of 10 equally spaced 100 mm flat plates (vanes). The airflow will continue downstream, mostly uninterrupted when the vanes are parallel to the wind tunnel floor; the airflow will be blocked when the vanes are perpendicular (vertically orientated) and redirected at an angle when the vanes are slightly rotated.

The gust-wind generator device, with specifically developed software and hardware is capable of reproducing the gust-wind phenomenon that is required and can easily be tuned by means of a few parameters. As depicted in Fig. 4, the vanes’ swing speed, angle, and frequency can be controlled by setting the control parameters in the control system. The redirection of the flow leads to an initial uniform flow oscillating up and down and produces an expected non-stationary flow.

Fig. 4

Control graphical interface

In order to achieve the goal of simulated gust wind, the vanes of the gust-wind generator device, starting from an initial position with a “closed state” 90° angle of attack, are rapidly angled downward to an “open state” as 0° angle of attack (i.e., counter-clockwise in Fig. 3). The transience of the flow (and thus the non-stationarity of the “mean”-wind velocity) is mimicked by returning the vanes to their original position and returning the flow to its most uninterrupted state. These actions produce a signal marked by periods of sudden temporal changes in wind velocity. Owing to the advantages of the actuation system, vanes can be successfully operated at a maximum initial value, corresponding to the configuration wherein the vanes are parallel to the floor with the device in its “open state.” Moreover, the system supports the usage of the generator in this configuration, enabling the vanes to reciprocate to swing up and down.

4 Experimental study of gust generator performance

In order to assess the performance of the gust-wind generator, experiments were conducted at the National Engineering Research Center for Construction Technology of High-Speed Railway at the railway campus of Central South University. The gust-wind generator was installed at the upstream high-speed section of the CSU wind tunnel to provide as much space for the flow and formation of gust wind as possible, as shown in Fig. 5. Vanes installed on the gust-wind generator were able to rotate through a full 90^o, with 0^o corresponding to the flat plate horizontal to the approaching flow and 90^o corresponding to the flat plate perpendicular to the oncoming flow, imposing up to 40% blockage of the initial flow region.

Fig. 5

Layout of gust-wind generator

4.1 Measurement devices

The system to measure the gust-wind velocities uses a Cobra Probe (Turbulent Flow Instrumentation), coupled with a pitot tube. The Cobra Probe is a four-pressure-hole probe with a high-frequency response capable of measuring velocities in three directions and mainly sensitive to flow variations normal to the probe with an accuracy of 0.5 m/s and a range of 0-60 m/s (Zou et al. 2022a, b, c), as shown in Fig. 6. Meanwhile, the probes are mounted on a traverse system, allowing measurements at different locations in space. By positioning the probe parallel to the free stream flow, the velocities induced by the gust-wind generator can be measured. The Cobra probe was placed downstream of the wind tunnel and behind the gust-wind generator to contribute to the development of oncoming flow. In this study, measurements are conducted at the center of the test section after a series of tests done for a range of flow speeds, actuation frequencies, vane deflection amplitudes and locations to provide suitable gust-wind results.

Fig. 6

4-hole cobra probe manufactured and calibrated by Turbulent Flow Instruments (http://www.turbulentflow.com.au/)

4.2 Data processing

The measurements with the Cobra probe are recorded over 30 seconds at a sampling frequency of 2 k Hz. Due to the initial measurements having a low signal-to-noise ratio, three techniques for obtaining the time-varying, non-stationary (slowly-varying) mean velocity, U, of the gust-wind flow are first compared and analyzed: (1) moving average (MA), and (2) decomposition using discrete wavelet transform (DWT).

The “moving average” is characterized by less computation and high significance, which is a more popular approach that has been utilized by previous studies to filter measurement records and extract the “mean” velocity from the time-varying, non-stationary winds (Yang et al. 2004; Belu and Koracin 2013). Recently, the other method of discrete wavelet transformation (DWT) has been proposed for the non-stationary “mean” extraction of time-varying flow velocity data (Kijewski and Kareem 2003; Wang and Kareem 2004). The algorithm of DWT operates by decomposing a signal into an approximation component that captures low-frequency contents contributing to the time-varying mean and a detail component that contains high-frequency contents. A series of high and low pass filters were used to progressively find the wavelet coefficients from the highest level to the mean value level. The reliability of the DWT in the identification and characterization of non-stationary random signals was verified and has been applied in ocean, wind and earthquake engineering (Gurley and Kareem 1999).

4.3 A sudden change in wind velocity

A close look at the literature reveals that the main properties of mountain-valley winds must be achieved to generate realistic non-stationary gusts. The two primary factors considered were the magnitude of the gust wind speed and the time scale of the velocity increase. For mountain area-related gust winds, Simiu and Scanlan (1996) summarize ranges of gust wind magnitudes afrom 3 m/s to 30 m/s and gust wind durations from a few minutes to 20 min or more. Hence, the average speed of wind mutation is one of the characteristics of gust wind, and the average wind velocities undergo at least one change in a short period. In order to highlight this characteristic of gust wind, the status of wind speeds before and after changes are defined as U₁ and U₂, and a time of radical changes in wind velocities is expressed as Δt = T₂ − T₁ (T₁ is the ending point of U₁ and T₂ is the beginning point of U₂), as shown in Fig. 7. To generate the sudden-change airflow, the wind tunnel was first started up until the wind velocity U₁ reached the stable state, after that, the wind velocity will be changed U₂ in the shortest period Δt until it reached the stable state again.

Fig. 7

A sudden change of wind velocity

Figure 8 shows the 2 sample decomposition of wind speed records simulated by the gust-wind generator. The overlay of the total time history of the sudden change flow record confirms that interdependently of the choice of averaging methods, the computed non-stationary mean follows the temporal trend of the velocity. The “mean” values of wind velocity vary significantly with time. Both approaches display similar changing rules of mean-wind velocities with time, and the local peaks vary between 1.40% and 6.61% percentage difference, the peaks from the DWT approach tend to be slightly larger in magnitude by 0.24% than those from the running mean.

Fig. 8

Wind speed time series sample

Figure 8a shows that the wind velocity accelerated from 2.31 m/s to 20.23 m/s and when an action potential is initiated, the gust vanes go from the closed state to the open state. The airflow can pass through in an instant and the accelerating time Δt was defined as 0.33 s, and the corresponding acceleration was 54.30 m/s². Meanwhile, the gust-wind generator can negatively control operation, as shown in Fig. 8b, the wind velocity decelerated from 19.66 m/s to 1.76 m/s and when an action potential is initiated the gust vanes go from 0^o to 90 ^o. The airflow stopped in an instant and the decelerating time Δt was defined as 0.44 s, and the corresponding acceleration was 40.68 m/s².

4.4 Continuous gust

For a sinusoidally oscillating flow, the continuous sinusoidal mode of the gust wind generator was tested for a range of amplitudes and frequencies for two tunnel-free-stream velocities. The initial velocity in a wind tunnel is U₁ = 12 and U₁ = 25, respectively. The time histories in Fig. 9 indicate the reciprocating characteristic of the oncoming flow. As expected, the amplitude of the induced gust is increases with larger gust vane deflections. The oscillatory pattern is visible for all combinations of amplitude and frequency, however at higher frequencies for the low wind speed amplitude case, the pattern becomes less sinusoidal with distorted peaks. Pure sinusoidal waves, especially for vertical turbulence, can hardly be generated because the attenuation of turbulence along the test section and the wind tunnel wall interference are complex. However, the sinusoidal type transient gust is another type of unconventional wind flow, which is normally observed in mountain areas and cannot be generated in a traditional boundary layer wind tunnel, as mentioned above in Fig. 1. By comparison in the relation curves Figs. 1 and 9, wind speed varies with time generated in the CSU wind tunnel are similar to that obtained in a mountain area.

Fig. 9

Time history of the wind velocity of a single pseudo harmonic

5 Conclusion

A new gust-wind generator at the Central South University (CSU) wind tunnel is utilized to simulate gust winds in mountain areas. The system is worked through an advanced control system and a technology basis of theoretic to identify vanes movements leading to the target wind speed time history. The system is capable of generating applicable transient nature of wind speed and the current study simulated two kinds of typical mountain-valley wind. In the next step, the gust-wind generator will be improving and launching projects on other types of non-stationary winds and their effects on railway-related structures.