Wind tunnel analysis of the slipstream and wake of a high-speed train

Abstract

The slipstream of high-speed trains is investigated in a wind tunnel through velocity flow mapping in the wake and streamwise measurements with dynamic pressure probes. The flow mapping is used to explain the familiar slipstream characteristics of high-speed trains, specifically the largest slipstream velocities in the near wake. Further, the transient nature of the wake is explored through frequency and probability distribution analysis. The development of a wind tunnel methodology for slipstream assessment is presented and applied, comparing the output to full-scale results available in the literature. The influence of the modelling ballast and rail or a flat ground configuration on the wake structure and corresponding slipstream results are also presented.

Introduction

The slipstream of a high-speed train (HST) continues to be an important aspect of aerodynamic performance and safe operation. Slipstream is the air flow induced by the train׳s movement as experienced by a stationary observer. Such flows can be hazardous to waiting commuters at platforms and track-side workers (Pope, 2007). The flows can also cause damage to track-side infrastructure. Regulations are currently in place aimed at reducing risks for HSTs; for example, the European Railway Agency׳s (2008) (ERA) Technical Specifications for Interoperability (TSI) and the European industry norms outlined by the European Committee for Standardization (EN) (CEN European Standard, 2009). Thus, there is a need for both improvement in the understanding of the slipstream of HSTs and methodologies for testing HSTs in the development stages.

The slipstream of a HST under ‘standard operation and configuration’, defined here as a single train with one nose and one tail travelling on a straight track over flat ground with no crosswind present, is investigated. This idealised train is modelled to isolate the slipstream characteristics generated by the train׳s essential generic geometry in an ideal environment. Under these conditions, the slipstream of a HST has a local peak velocity at the nose passing, a gradual increase in velocity as the boundary layer develops along the length of the train, followed by the largest peak in the near wake of the vehicle (Baker, 2010, Baker et al., 2012a, Baker et al., 2012b, Baker et al., 2001). These slipstream characteristics correspond to the description by Baker (2010) of flow around a HST having three distinct regions: nose, boundary layer and wake regions. These general characteristics of a HST׳s slipstream, illustrated in Fig. 1, are referred to herein as the ‘standard slipstream profile’ and have been found by a number of researchers in full-scale track-side experiments (Baker, 2010, Baker et al., 2012a, Baker et al., 2012b, Sterling et al., 2008).

Inter-carriage gaps have been found to cause perturbations to this general description as peaks, troughs or waves (Muld et al., 2013b, Pii et al., 2014, Bell et al., 2014), however these do not appear to significantly change the rate of increase of the boundary layer thickness. A local tail peak has also been identified in a number of HST slipstream profiles in full-scale experiments (Baker et al., 2012a), scaled experiments (Gilbert et al., 2013), and numerical simulations (Muld et al., 2013b, Hemida et al., 2013), but is not a standard feature and is likely dependent on geometry and measurement position. Both features are included in Fig. 1 as dotted lines to indicate that they are not standard, nor the focus of this research. Further, as the flow around HSTs is highly three dimensional, the slipstream profile as measured by a single streamwise line, as indicated in Fig. 1, is highly sensitive to measurement position, with the shape of the slipstream profile – even the relative magnitudes of the peaks – being susceptible to changes. However, in general the slipstream velocity decreases with increasing height above ground and distance away from train, as shown in full-scale experiments (Sterling et al., 2008) and numerical simulations (Hemida et al., 2013).

A number of known flow mechanisms can be identified in the wake of a high-speed train: shear layers, von Kármán-type vortex shedding, separation and recirculation regions and a pair of twin counter-rotating longitudinal vortices (Morel, 1980, Weise et al., 2006, Muld et al., 2012). The contribution of twin counter-rotating vortices to wake topology has been identified as a particularly important feature in characterising slipstream (Baker, 2010, Weise et al., 2006, Muld et al., 2012). These vortices move downwards and outwards due to the mutual induction and interaction with the ground as they progress away from the vehicle (Weise et al., 2006, Muld et al., 2012, Heine et al., 2013, Schulte-Werning et al., 2001, Yao et al., 2013), with some researchers predicting that they exhibit spanwise oscillations (Muld et al., 2012, Yao et al., 2013, Schulte-Werning et al., 2003).

The counter-rotating vortices are created by the interaction between the down-wash over the roof and tail of the train and the flow around the sides of the train in the transition from a constant cross-section to the end of the tail. Other bluff body ground vehicles such as cars also contain a pair of counter-rotating vortices in the wake, and have been researched extensively primarily motivated by drag reduction. Such research, such as the widely referenced `Ahmed body׳ (Ahmed, 1983), is not necessarily due to the uniqueness of HST geometry; slender (high length:height), small aspect ratio (height:width ratio $\approx 1$) and streamlined nose/tail with no fixed separation points.

The link between a HST׳s wake topology and its slipstream has not been explicitly established in the literature. Further, full-scale experiments (Baker et al., 2012a, Baker et al., 2012b, Sterling et al., 2008), scaled experiments (Baker, 2010, Bell et al., 2014) and numerical investigations (Muld et al., 2013b, Pii et al., 2014, Hemida et al., 2013) have found high run-to-run variance, which suggests that understanding the transient wake in addition to the time averaged wake is necessary to understand slipstream. The work presented aims to bridge this gap.

Insight into the time-averaged wake and its transient features is achieved by utilising the train-fixed frame-of-reference in a wind tunnel methodology to perform time-averaged flow mapping and point-wise frequency and probability distribution analysis. This is the primary benefit of the wind tunnel methodology, the ability to map the flow. A wind tunnel methodology does however have a number of limitations. The train-fixed frame-of-reference does not easily allow for gust analysis, currently used to measure a HST׳s slipstream performance in the typical sense, as individual runs are not captured. Further, the presence of a stationary floor in wind tunnels rather than using a rolling road or suction results in a boundary layer developing on the floor, which does not occur in full-scale operation.

The presence of a stationary floor is a common experimental limitation for many wind tunnel experiments and one that is only mitigated with difficult and costly solutions such as boundary layer suction or a moving floor (rolling road). Experiments by Kwon et al. (2001) highlight the difficulty of such solutions to HST investigations due to their long bodies, having to apply suction at multiple slot locations with great care, noting that their position could influence drag measurements. Further, inclusion of a track and ballast shoulder ground configuration, as discussed below, prevents such boundary layer treatments.

The effect of a stationary floor on the aerodynamics of ground vehicles has been investigated by a number of researchers, with primary concern on the influence on drag and lift prediction (Choi et al., 2014). No definitive effect has been established, with the magnitude and even direction of the effect on lift and drag varying with geometry and distance to the ground (Choi et al., 2014). Vehicles with rear diffuser-type geometry have been found to be most susceptible to the influence of floor motion (Bearman et al., 1988). Some researchers have investigated the influence floor motion has on the near-wake structure. Scaled experiments by Bearman et al. (1988) and Strachan et al. (1988), and numerical simulations by Krajnović and Davidson (2005) have found that the pair of counter-rotating longitudinal vortices, which develop over the C-pillar of generic automotive vehicles – structures similar to that dominating the wake of the HST presented in this work – were largely insensitive to the stationary floor in the near wake.

Large Eddy Simulations (LES) of Krajnović and Davidson (2005) at a Reynolds number of 2×10⁵ indicated that the stationary floor influenced the base pressure, and reduced the clarity of dominant frequencies found in the transient flow. Despite this, velocity profiles at multiple distances in the near wake displayed only minor differences between a moving and stationary floor in close proximity to the ground, leading the authors to conclude that the longitudinal vortices were relatively insensitive to ground effect. The scaled experiments by Bearman et al. (1988) presented velocity profiles from pulsed-wire anemometry together with drag and lift measurements which indicate that the effect of a moving floor was negligible for a Davis generic automotive model with zero upsweep angle. Similarly, Strachan et al. (1988) noted that the C-pillar vortices of an Ahmed body were insensitive to the motion of the floor in their scaled experiments at a Reynolds number of 1.7×10⁶ (based on length) with a rolling road based on velocity profiles obtained from Laser Doppler Anemometry (LDA). Kwon et al. (2001) in scaled wind-tunnel experiments identified differences in oil-flow visualisation on the floor with and without tangential blowing, with a more coherent pattern and greater spanwise dispersion of the wake with tangential blowing. However, of course, it is expected that any differences in the wake would be greatest on the floor surface, and although outright drag coefficients exhibited minor differences for the same model between moving floor, tangential blowing and stationary cases, the change in drag coefficient was consistent across all floor treatments for a number of changes to geometry.

Scaled model experiments to assess slipstream have been performed at moving model facilities (Baker, 2010, Bell et al., 2014, Gilbert et al., 2013). The moving model results in the correct relative motion between the ground and the model, thus no ground boundary layer develops as in stationary floor wind tunnel experiments. This methodology also has the benefit of the ground-fixed frame-of-reference as in full-scale experiments, allowing gust analysis of individual runs to be performed. A moving model experiment also has limitations, where approximately 20 (CEN European Standard, 2009) runs are required to obtain statistically stable measurements in transient areas, such as the near wake of a HST. Flow mapping in the near wake, within the path of the vehicle is also very difficult and has only recently been proved possible using high speed Particle Image Velocimetry (PIV) by Heine et al. (2013).

Two ground configurations: ‘true flat ground’ (FG) and ‘single track ballast and rail’ (STBR) are common in experimental and numerical investigations. Ground configuration is not specified in the EN for scaled model slipstream experiments, however STBR was required for head pressure pulse investigations in the 2009 EN (CEN European Standard, 2009) – the 2013 EN (CEN European Standard, 2013) revision excluded the rails from this configuration – and crosswind investigations (CEN European Standard, 2010). The STBR ground configuration has also been proposed to be better suited to stationary floor experiments as it potentially lifts the model out of the ground boundary layer (Schober et al., 2010). The latter proposed benefit of using STBR assumes that the boundary layer does not develop over the STBR to the same extent as for a flat floor. The sensitivity of slipstream results to a STBR or FG configuration was investigated and is presented in this work.

The reduced length-to-height ratio (L/H) is necessary to allow as large a scale model in a test section – in order to achieve high Reynolds numbers. This is also a limitation of moving model experiments for similar practical reasons. The high L/H of HSTs also presents a challenge numerically as although technically they are not limited in space, computational resources become an issue, with increased L/H coming at the cost of Reynolds number and grid refinement. Although acknowledged in the literature (Weise et al., 2006), L/H has only recently been investigated by Muld et al. (2013a), and the effect of L/H has yet to be explicitly quantified.

This work, part of a collaboration between Monash University and Bombardier Transportation, aims to investigate wind tunnel testing and analysis techniques for use as a methodology for assessing the slipstream of HSTs. If an analysis technique can be developed for translating the data and predictions from wind tunnel measurements to standard train reference frame the authors believe that wind tunnel testing will offer significant benefits complementing other methodologies, particularly during the prototype phase of HST development. The experimental investigation presented includes streamwise measurements with dynamic pressure probes, to which the first iteration of a method for performing gust analysis in the manner specified by the TSI regulations, in the train-fixed frame-of-reference, has been applied. The presented flow mapping in the wake, also with dynamic pressure probes, aids in the identification and quantification of the flow mechanisms responsible for the transient characteristics which affect TSI-type analysis and thereby improve the understanding and predictive capabilities of the slipstream and wake structure of high-speed trains.