Tracking of Bristle Tip Deflections to Demonstrate Blow-Down in Brush Seals

Gas turbines are an energy-dense solution for aircraft propulsion and power generation, with a wide range of power outputs and fuel versatility that offer advantages over other technologies in the aviation and energy sectors. The secondary air system in a gas turbine diverts air from the main gas path to cool high-temperature turbine hardware and seal bearing and disk cavities. Sealing is crucial in maintaining the efficiency of the system by regulating leakage flows between the interface clearances that exist between rotating and stationary components. Seals therefore have a significant impact on the overall performance of gas turbines [1,2] and are commonplace in a typical engine design, as shown in Fig. 1. Improving seal technology is hence critical to achieving the targets set by the International Air Transport Association Fly Net Zero by 2050 and the International Energy Association Net Zero by 2050. Sufficient reduction of global greenhouse gas emissions to limit temperature increases will require enhanced efficiency of turbine technologies. Improved effectiveness of the secondary air system, achieved in part through enhanced sealing performance, facilitates higher turbine temperatures and reduced specific fuel consumption.

Labyrinth seals are widely used in turbomachinery because they offer a robust and low-cost sealing solution [4]. Energy is dissipated through a series of cascading cavities [2], or teeth, which are shown insert in Fig. 1. Sealing performance is heavily dependent on the clearance between the seal teeth and the adjacent rotor surface. Tighter clearances pass smaller leakage flows but incur a higher probability of rotor–stator contact during rotor excursions, causing seal wear and a reduction in operational life.

Tight clearances can be offered by brush seals without compromising on operational life as they are a contact-compliant alternative to labyrinth seals. Also shown in Fig. 1, a brush seal cross-section is defined in Fig. 2(a), with key features highlighted. A bristle pack of closely arranged flexible fine wire bristles that offer resistance to the oncoming flow are commonly welded at the root and protrude radially inwards toward the rotor surface. An alternative design where the bristle pack is clamped at the root is also widely applied.

Brush seals can be assembled with clearances between the bristle tips and rotor surface, as shown in Fig. 2, but also in a line-on-line (LoL) condition, whereby the unloaded bristle tips are initially in contact with the rotor surface. These bristles are assembled at a lay angle to the radius (see Fig. 3) permitting shaft rotation in the desired direction and ensuring that they bend to accommodate rotor excursions without buckling, protecting the seal and rotor from permanent wear and performance degradation. Brush seals can therefore be assembled with tighter clearances that result in the leakage flow passed reducing by approximately 80% [2] relative to labyrinth seals, driving substantial turbine efficiency gains. Further benefits include a lower mass and reduced axial space [5] that ease assembly and maintenance.

The front plate encloses the bristle pack on the upstream side, protecting the bristle pack from the oncoming high-pressure flow and minimizing the risk of damage during installation [6]. The back plate provides structural support on the downstream side, helping the bristle pack to accommodate large pressure loads.

Figure 2(b) defines salient dimensions in brush seals and shows a pressure relieving (PR) back plate design. The PR pockets represent an alternative design to the flat-surfaced, conventional back plate configuration which aims to reduce the effect of high seal stiffness on durability [7]. One example of this is hysteresis, where the bristles fail to recover their positions following a rotor excursion that displaces the pack, thereby increasing the clearance area and enlarging leakage flow [8]. Relatively large pressures build in the pockets incorporated into the back plate surface which offset the axial load, reducing frictional forces within the bristle pack and against the back plate surface [7].

When a pressure load is applied to the brush seal, the bristles deflect axially downstream and compact together against the back plate surface, pivoting about the bristle root. This behavior is illustrated in Fig. 4, which contrasts the unloaded and loaded bristle pack. As the bristles compact together, the gaps between them that constitute the leakage pathways through the bristle pack close and the pack becomes less porous, as identified by the photographs presented in Fig. 5. The oncoming flow also imparts a bending moment onto the bristles due to their lay angle, causing the bristles to deflect toward the rotor surface and to close the clearance [2,9]. This radial deflection—known as blow-down—is also highlighted in Fig. 4. This behavior ensures that performance gains are maintained during rotor excursions and as the bristles wear. A bending moment is similarly generated in the tangential plane, meaning that the bristles deflect in all three spatial axes when subjected to a pressure load.

Due to the reduced frictional forces, increased blow-down is expected to occur in PR brush seals. While this can allow for enhanced sealing performance, it can also result in premature bristle wear [7]. The impact of PR back plate designs and the complex nature of bristle tip deflections remain topics of limited study.

Typical brush seals for aero-engine applications are of small scale and exist in close proximity to high-speed rotating hardware, engendering a difficult environment in which to investigate the fluid dynamic and mechanical behavior. This study utilizes a linear, large brush seal of ten-times scale, referred to as the TEN-X seal, that was previously employed by Bowen et al. [3,10] to characterize the inter-bristle pressure field and leakage flows. A novel application of Digital Image Correlation (DIC) technology tracked the individual bristle tips across a section of the pack to directly measure axial, radial, and tangential deflections for a PR brush seal at two clearance conditions. Analysis of tip deflections in conjunction with leakage characteristics and the inter-bristle pressure field is carried out to investigate the behavior of PR brush seals. This paper presents the first direct, experimental measurements of blow-down throughout the bristle pack. Results published here provide a unique insight into the mechanical characteristics of brush seals and the impact of bristle deflections on the fluid dynamic behavior.

This section presents a brief review of brush seal research. Publications detailing experimental and theoretical investigations are included, as well as papers that provide reviews of existing literature. In Secs. 2.1 and 2.2, research focused on conventional and pressure relieving brush seals is discussed. Finally, in Sec. 2.3, studies of bristle deflection are highlighted.

A comprehensive overview of turbomachinery sealing solutions is provided by Chupp et al. [4], in which labyrinth and brush seals are covered in detail. Aslan-Zada et al. [2] present an extensive review of published literature, specifically focusing on the application of brush seals as an alternative to labyrinth seals. Along with many other studies, they highlight the sealing gains achieved by brush seals, but acknowledge the requirement for further experimental analysis to improve the understanding of brush seal fluid dynamic behavior.

Bayley and Long [11], Chen [12], and Chen et al. [13] conducted combined experimental and theoretical investigations into the fluid dynamic behavior of brush seals. Bayley and Long [11] made measurements of the pressure field from a non-rotating circular brush seal with a small interference. Because this seal was of realistic aero-engine operational scale, insight was limited to pressure distributions on the back plate and rotor surfaces, lacking the necessary spatial resolution to examine the inter-bristle pressure field. Studies at the University of Oxford [12,13] presented data collected using a linear five-times scale brush seal in which a selection of solid bristles were replaced by hollow tubes, allowing for interrogation of the inter-bristle pressure field. These publications provide limited insight, studying only a small range of operating conditions on brush seals of an obsolete design. In contrast, the study presented here uses a state-of-the-art aerospace brush seal design that was tested over a wide range of operating conditions.

Measurements of static pressure within the bristle pack were collected from the linear, ten-times scale brush seal (TEN-X) at the University of Bath. Bowen et al. [10] published results from this facility characterizing the leakage characteristics of line-on-line and clearance brush seals, at a wide range of operating conditions, in direct relation to the inter-bristle pressure field. Changes in bristle blow-down were shown to impact the sealing performance and the distributions of pressure more significantly than pack compaction, highlighting the need for an enhanced understanding of this behavior.

A numerical investigation of brush seal pressure and flow fields was carried out by Doğu and Aksit [14] for different back plate configurations. Line-on-line and clearance conditions were considered. A baseline conventional design was studied in comparison to alternatives including two separate PR arrangements, which contained single and multiple constant pressure pockets within the back plate surface, respectively. The bulk porous medium approach for the bristle pack presented by Doğu [15] was utilized to evaluate leakage characteristics as well as the pressure and flow fields, using axial and radial velocity vectors and turbulent kinetic energy distributions. This analysis concluded that back plate geometry had limited impact on seal leakage but did allow for control over the back plate surface pressure field, which influenced the resulting bristle deflection, wear, and heat generation.

The sealing performance of PR brush seals relative to conventional designs was evaluated by Pekris et al. [7]. Measurements of leakage, torque, and rotor temperature were taken from a multi-seal test rig that housed a clearance brush seal at a range of pressure loads and shaft rotational speeds. The use of a PR back plate resulted in a small leakage penalty but reduced hysteresis effects and caused a significant drop in rotor torque. For the seals tested, only the use of an active PR back plate design, in which pockets on the back plate surface were manually pressurized, enhanced sealing performance relative to a conventional brush seal. Such active pressure designs further reduced inter-bristle and bristle-back plate friction but due to the increased blow-down, the potential for wear was highlighted.

Following on from their previous work, Bowen et al. [3] evaluated the performance of line-on-line brush seals with conventional and PR back plate designs. This study employed a concomitant methodology that exploited the benefits of the TEN-X seal in addition to an engine-representative, rotating brush seal. Larger seal leakage for both PR brush seals tested at line-on-line clearance conditions was attributed to the impact of back plate geometry on the inter-bristle pressure field. However, it was acknowledged that different conclusions would likely be drawn when investigating clearance conditions, which form a key part of the focus of this study.

A broad scope of literature exists on the subject of bristle deflections and relevant phenomena such as hysteresis; however, displacements have never been directly measured throughout the bristle pack.

Franceschini et al. [9] conducted wind tunnel tests and numerical simulations which identified aerodynamic mechanisms that drive blow-down, such as radial distributions of pressure throughout the bristle pack and swirl in the clearance region. Chen [12] presents axial bristle displacements inferred by matching static pressure measurements in the deflected bristle tips to those on the rotor surface, shown in Fig. 6. Experimentation at Cross Manufacturing Ltd. demonstrated blow-down through analysis of leakage characteristics and rotor torque measurements [16,17].

Optical measurements of bristle deflection have previously been employed by Schwarz et al. [18] and Bowen et al. [10]. Experimental results were presented by Schwarz et al. [18] from a circular brush seal of typical aero-engine scale with a clearance configuration that was not subjected to shaft rotation. The impact of pressure load on blow-down in the final, most downstream bristle row and overall pack thickness was measured. However, the small-scale of the studied seal did not allow for the high-fidelity analysis of tracking individual bristles throughout the entire width of the pack. Limited information was provided concerning the experimental methodology, calibration procedures, and analysis of uncertainties. The main focus of the study concerned the influence of the axial inclination parameter for clamped brush seals. Optical analysis utilized by Bowen et al. [10] was a simplistic approach that only considered axial bristle deflections. It did not allow for a large number of bristles to be studied or for displacements in the radial plane to be interrogated.

Theoretical studies including Doğu et al. [19] and Thorat and Bauer [20] estimated bristle blow-down by calibrating the clearance in combined orifice and porous medium models to match measured datasets.

Where blow-down results have been presented, the measurement techniques have been unique to each study, often indirect, and of low-fidelity. Deflections inferred from other results such as leakage characteristics and pressure fields are subject to large uncertainties, such as those in Fig. 6. Additionally, the variety of approaches has yielded inconsistency in the resulting conclusions. Where direct measurements have been made, limited insight has been gathered. Further study is required to form a complete characterization of the influence of seal clearance, back plate design, and operating conditions on bristle deflections in all three spatial axes. By interrogating bristle tip deflection in three spatial axes through advanced optical analysis, this study presents the first published directly measured results of bristle blow-down in brush seals throughout the bristle pack.

The Large-Scale Test Facility at the University of Bath consists of a linear brush seal of a ten-times scale, referred to as the TEN-X. Its large scale and linear geometry allow for the inter-bristle pressure field to be measured to a greater spatial resolution compared to true-engine scale, rotational test facilities. The facility is described by Bowen et al. [10], who highlight the design considerations made to ensure representative brush seal behavior. This was demonstrated experimentally and theoretically by Bowen et al. [3] in comparison to a typical-scale aero-engine brush seal that was designed and manufactured following the in-house processes at Cross Manufacturing Ltd. It was demonstrated that insight gathered from the TEN-X was generically applicable to brush seals. Technical data for the TEN-X seal are provided in Table 1, where geometric definitions for key parameters are shown in Fig. 2 for both conventional and pressure relieving back plate configurations. A cross-section view of the Large-Scale Test Facility is given in Fig. 7, which highlights the air pathway as well as the upstream and downstream static pressure measurement locations.

Twelve rows of tightly bundled solid stainless steel 304 hypodermic tubes of 1.02 mm diameter make up the bristle pack of the TEN-X seal. These bristles are assembled at a lay angle of 45 deg. Utilizing the in-house design tool at Cross Manufacturing Ltd., considerations were made to ensure geometric and physical similarity to a typical aero-engine brush seal, resulting in the bristle diameter and pack stiffness being scaled by a factor of ten.

Static pressure measurements were acquired on the back plate surface at the locations indicated in Fig. 2 and defined in Table 2. Measurements were also made at all 12 rows in the bristle pack at consistent equivalent radial—herein referred to as radial—locations to those on the back plate surface, as well as at the bristle tips. The hollow bristle tips that represent static pressure measurement locations at this position are visible in Fig. 5. All static pressure recordings were determined using HMA Series Amplified Pressure Sensors with a National Instruments (Austin, TX) CompactDAQ data acquisition system and LabVIEW analysis software. Compressed air was supplied to the TEN-X seal with the leakage flow measured using a Bronkhorst (Ruurlo, The Netherlands) F-106EZ mass flowmeter of range 0.02–1.0 kg/s. Flow temperature was measured in the inlet pipe downstream of the mass flowmeter using a K-Type thermocouple.

Bowen et al. [3] present an uncertainty analysis of static pressure and leakage flow measurements made using the Large-Scale Test Facility with the TEN-X brush seal.

For all tests reported here, static pressure in the downstream chamber remained atmospheric, with the flow exhausting to the surroundings through the outlet. The inlet mass flow rate was increased from zero, yielding a rise in $pu$ and $Δ$p.

The TEN-X seal was tested with conventional and PR back plate configurations, as shown in Figs. 2(a) and 2(b), respectively. The PR back plate contains three pockets of a depth of 2.5 mm that encompass the majority of the back plate surface area in order to reduce the contact area between the bristle pack and the back plate surface. The size of each pocket increases with radial distance from the rotor surface.

The TEN-X brush seal can be assembled into the Large-Scale Test Facility to achieve a variation in seal clearance. Measurements were collected in this study for the clearance conditions defined in Table 3 for both back plate configurations. A LoL setup is for zero clearance where the unloaded bristles are in contact with the rotor surface. Clearances C1 and C2 are of increasing magnitudes. Additional measurements from the TEN-X seal by Bowen et al. [10] at different clearance conditions are shown in Sec. 5 for comparison.

Results are presented in this study in terms of normalized parameters that allow for comparison between seals of different design configurations and different test conditions.

This study uniquely measures the magnitude of deflection in three spatial axes for a large number of bristle tips, employing DIC technology.

The DIC apparatus utilized two LaVISION (Göttingen, Germany) Imager M-Lite 5M cameras, which have a high resolution of 2.4k $×$ 2.0k pixels and a frame rate of 50 Hz. These were separated by an angle of 30 deg to view the bristle tips through an optical access window, as shown in Figs. 8(a) and 8(b). The images provided by each of the two cameras were combined through triangulation to produce the perspective indicated in Figs. 8(c) and 8(d), allowing for data in all three spatial axes to be measured. Note, only four of the 12 bristle rows are shown here for brevity.

A section of the bristle tips of the TEN-X seal was painted with a layer of white emulsion and overlayed with a random black speckle pattern. This resulted in each bristle tip in the studied region having a unique and individual speckle pattern that was easily visible and identifiable on top of the contrasting background.

Images were taken of the bristle tips under unloaded and loaded conditions, during which each bristle tip would deflect radially, axially, and tangentially, as indicated by Fig. 8. Post-processing of the images was carried out using LaVISION davis software. In-software target points were located manually onto the bristle tips in images of the unloaded pack to identify the speckle pattern on individual bristles. These target points tracked the corresponding bristle in subsequent images of the deflection procedure through the identification of the appropriate speckle pattern. This process is illustrated by the schematic diagram in Fig. 9, in which the colored target points track their assigned bristle tip as it deflects in three spatial axes. The automated target point tracking system allowed for the analysis of a high number of bristles across a range of test cases. This process was calculated to incur an average error of 0.14% compared to manually located gauges in terms of the displacement of the target point from the undeflected bristle tip.

The DIC system was calibrated using a micro-calibration plate fixed to the bristle tips at multiple angles to the axial and tangential axes. Following this process and with triangulation of the images from the two cameras, the target points provided positional data in all three spatial axes. The target points located the three-dimensional position of each bristle tip studied from images of the deflected and undeflected pack. Therefore, displacements in all three axes were calculated relative to the undeflected bristle. The calibration ensured that the measured displacements contained an associated error of <0.2 pixels, which corresponds to a focus area contained within 0.69 $μ$m$×$ 0.69 $μ$m.

In the interrogated region of the bristle pack, 152 bristle tips were analyzed, accounting for 12 or 13 in each row. This value varies due to the hexagonal packing of the bristles, as indicated in Fig. 9. Measurements were taken for every bristle in this studied region for all 12 rows so that the individual bristle movements and generic bristle row deflection behavior could be examined.

At high-pressure loads, some bristles were subjected to high-frequency fluttering motion that exceeded the frame rate of the cameras. The resulting images showed these bristles out of focus, and therefore the automated target point could not identify the correct bristle. Such measurements were identified as errors through statistical filtering and were not included in the average bristle row displacements shown in Secs. 4 and 5. The uncertainties of bristle deflection results presented throughout this study, accounting for calibration and averaging errors, are evaluated in Appendix  B.

In this section, bristle tip deflections are examined for the TEN-X seal at the clearance conditions C1 and C2 (defined in Table 3) with a pressure relieving back plate configuration. Independent bristle and row-averaged deflections are first investigated at clearance C1, with the impact of pressure load and axial location on bristle displacements in the axial and radial planes studied. Subsequently, comparisons are made with measurements taken at clearance C2 to demonstrate the impact of seal clearance.

Using the convention established by Bowen et al. [3], hollow symbols are used to represent data collected from a PR brush seal. Circular symbols refer to measurements at clearance C1, and triangular symbols correspond to data taken at the larger clearance C2. Each figure presented in this section distinctly defines symbol color, which refers to an individual bristle, bristle row, or applied pressure load in specific figures. An accompanying silhouette in each figure specifies the direction (or directions) of bristle deflections presented, as well as the region of the bristle pack investigated, where appropriate.

Measurements of radial bristle deflection (blow-down) are shown in Fig. 10 for each individual studied bristle tangentially across Row 1, as defined in Fig. 9. The 13 studied bristles are all located in the center of the test section, minimizing the impact of end effects from the linear brush seal. As for all results in this subsection, the clearance condition C1 and a PR back plate are used. Greater pressure loads are shown to linearly correlate to increased blow-down and hence reduction of the clearance area between the bristle tips in Row 1 and the rotor surface.

Results can be seen to follow a common trend with no outliers, demonstrating that an averaged value can be used for row-to-row comparisons. However, variations in blow-down displacements were observed between individual bristles. Individual bristles are unique and are subject to manufacturing tolerances. Furthermore, inter-bristle friction varies throughout the row. Figure 10 highlights the novelty of the application of DIC technology, in which individual and bulk-averaged bristle deflections can be evaluated. The variation of radial bristle deflections is further analyzed in Appendix  A.

Figure 11 shows the row-averaged blow-down measurements over a range of applied pressure loads for each of the 12 rows in the axial direction, as defined in Fig. 9. The magnitude of blow-down consistently decreases in subsequent downstream rows, from an average of approximately 0.75$cm$ in Row 1 at $Δ$p = 4.4 bar to approximately 0.4$cm$ in Rows 11 and 12 for the same pressure load. The changes in blow-down with pressure load also decrease in subsequent downstream rows. Deflections in the axial and tangential planes in the upstream rows were also observed to be larger than those downstream, in agreement with findings from Chen [12]. This trend is attributed to the increased reaction and frictional forces experienced by the downstream bristles due to their proximity to the back plate.

The correlation between row-averaged bristle tip deflection in the axial and radial planes is presented in Fig. 12. Once again, the symbol color defines the row location. For all but the last bristle rows, a non-linear trend exists between measured axial and radial tip deflections, which becomes linear at larger magnitudes. Conversely, Rows 11 and 12 experience a linear relationship between axial and radial tip displacements throughout. As shown in Fig. 5(a), the bristle tips in an unloaded pack are splayed, so inter-bristle reaction and frictional forces are relatively low. When subjected to a small pressure load, the bristle pack compacts together against the back plate surface (Fig. 5(b)), increasing the resistance to further deflection for all bristles. Deflection behavior can therefore be split into two distinct categories: before pack compaction (non-linear) and after pack compaction (linear). The effect of initial compaction is reduced for bristle rows further downstream which have less remaining pack to compact into, and is negligible for the final row. This effect can also be noted in Fig. 11 by the initial sharp increase in radial deflection for the upstream bristle rows, which then progressively linearize for measurements where $Δ$p > 0.3. Meanwhile, for Rows 11 and 12, the variation in trends before and after the first non-zero measurement is significantly smaller.

The link between axial and radial tip deflections is also demonstrated in Fig. 13, in which the applied pressure load is now highlighted by the symbol color. Although not shown here for brevity, similar trends also exist when comparing axial or radial deflections with those in the tangential plane. Consistent with Fig. 12, this indicates that bristle tip deflection in a given plane, in addition to geometric parameters and applied pressure load, governs displacement in another.

The high resolution and accuracy of the DIC methodology are demonstrated by the clear trends exhibited in Figs. 10–13, in which very few outliers are noted. This is particularly clear in contrast to the data shown in Fig. 6.

The bristle tip deflection in the radial (blow-down) and axial planes are displayed in Figs. 14 and 15, respectively, comparing measurements at clearances C1 and C2. In general, a larger seal clearance correlates to an increase in overall bristle deflection.

The increase in blow-down due to the larger clearance is similar for all rows shown in Fig. 14, which represent the front (upstream), middle, and rear (downstream) boundaries of the bristle pack. Conversely, the influence of seal clearance on axial tip displacement diminishes further downstream. Figure 15 depicts a significant increase in axial tip deflection in Row 1, but only a negligible increase with seal clearance is shown in Row 12. The effect of initial pack compaction on axial and radial tip displacements is also demonstrated for the larger clearance condition C2.

Significantly steeper gradients of blow-down with respect to pressure load (see Fig. 14) are experienced for the larger clearance C2. This trend is not repeated when studying axial deflection (Fig. 15), for which the gradients of each dataset are similar regardless of clearance condition or axial position.

In this section, the sealing performance of conventional and pressure relieving brush seals is discussed with reference to the inter-bristle pressure field and bristle deflection behavior. In Sec. 5.2, the influence of back plate configuration on bristle deflection is investigated, for which the impact on the seal leakage is discussed.

Solid symbols depict data obtained with a conventional back plate, while hollow symbols describe measurements made with a PR back plate. Square, circular, and triangular symbols refer to the LoL, C1, and C2 clearance conditions, respectively. Symbol color refers to the applied pressure load, which is specifically defined in each subsequent figure. Accompanying silhouettes specify the region of the bristle pack investigated.

Figure 16 shows the leakage performance of the TEN-X seal with conventional and PR back plate configurations at LoL, C1, and C2 clearance conditions. It was observed by Bowen et al. [3] that at the LoL condition, a PR brush seal causes a small increase in leakage flow. Conversely, at the large clearance condition C2, the PR back plate engenders a significant enhancement in sealing performance. At clearance C1, a very similar leakage performance is seen for the TEN-X seal independent of back plate design. The same trends are observed in terms of $ΦA$ and $ceff$ in Figs. 16(a) and 16(b), respectively.

Clearance condition C1 is hence considered to evaluate the influence of the PR design on leakage pathways. Figures 17 and 18 show the axial distributions of pressure through the bristle pack of the conventional and PR TEN-X seal at the bristle tips and at the radial location of Y = 0.76, respectively. In each figure, the locations of the static pressure measurements are highlighted in the accompanying silhouette of the seal cross-section. Consistent pressure loads are compared in each figure to isolate the influence of the back plate configuration at each radial location.

At the bristle tips (Fig. 17), the distributions of axial pressure are very similar for the conventional and PR back plates. Bowen et al. [3] showed that the pressure field at the bristle tips is relatively unaffected by the back plate configuration for the line-on-line brush seal, which is also shown to be true here for a clearance seal. In contrast, at the radially outermost location of Y = 0.76 (Fig. 18), the PR back plate significantly influences the axial distribution of pressure, as was also the case for a line-on-line brush seal [3]. The common, near-linear axial pressure drops observed in the front half of the bristle pack diverge for Z > 0.5 (Fig. 18), leading to significantly larger static pressures on the surface of the PR back plate than for the conventional configuration.

The leakage through a clearance brush seal can be separated into three key pathways, as illustrated in Fig. 19:

• $m˙1$⁠: the flow through the porous bristle pack

• $m˙2$⁠: the flow that travels radially inward down the back plate surface toward the fence height region

• $m˙3$⁠: the flow that travels axially through the clearance region

For a clearance brush seal, $m˙3$ constitutes the vast majority of leakage, with Doğu et al. [22] suggesting that this accounts for over 93% of all leakage, even for a small seal clearance. Such dominance means that the sealing behavior of large clearance brush seals was shown by Bowen et al. [10] to be akin to flow through an orifice and driven by pressure ratio.

As discussed by Bowen et al. [3], larger static pressures exist in the pockets of the PR back plate than on the surface of the conventional design (Fig. 18), causing an enhancement of leakage pathways $m˙1$ and $m˙2$⁠. This is attributed to increased pack porosity and a stronger radial pressure gradient down the back plate surface, which explains the amplified leakage passed by the PR brush seal in the line-on-line condition [3] (see Fig. 16), where the impact of $m˙3$ is removed.

The conventional and PR TEN-X seals pass a consistent magnitude of leakage at clearance C1 (Fig. 16), despite $m˙1$ and $m˙2$ being larger for the PR configuration. Therefore, it can be concluded that the PR brush seal must have limited the leakage pathway underneath the bristle tips (⁠$m˙3$⁠). This implies that this back plate configuration allows for increased bristle blow-down to close the clearance area. This conforms to the underlying theory described in Sec. 1.

The axial distributions of pressure presented in Figs. 17 and 18 for clearance C1 were consistent for C2, although not shown here for brevity. This demonstrates that the magnitudes of pathways $m˙1$ and $m˙2$ are similar for each back plate design, independent of seal clearance. Rises in leakage between clearances C1 and C2 are therefore driven by the enlargement of $m˙3$ because of the increased area underneath the bristle tips. The impact of bristle blow-down on sealing performance is hence heightened as the clearance increases. The greater radial deflection permitted by the PR back plate engenders the reduced leakage relative to the conventional brush seal at clearance C2 displayed in Fig. 16. Because the dominance of leakage pathway $m˙3$ amplifies at larger seal clearances, the scale of sealing improvement provided by the PR back plate increases further. It is worth noting that C1 and C2 represent smaller seal clearances than would be expected to occur during engine operation, in proportion to the bristle diameter. This means even more significant sealing benefits can be expected from the application of PR back plate configurations.

Figure 20 compares the measurements of axial bristle tip deflection from Row 1 of the PR TEN-X seal at clearance conditions C1 and C2 (presented in Fig. 15) with equivalent measurements taken by Bowen et al. [10] for the conventional TEN-X seal. The latter measurements were taken using a lower-fidelity method of optical analysis in which only one bristle per row was examined. An explanation of the methodology and an uncertainty analysis of the results is provided by Bowen et al. [10]. Figure 20 shows the change in measured axial tip deflection for different seal clearances; the clearances are not consistent between the conventional and PR datasets. However, consistent pressure loads are presented for each back plate configuration, isolating the influence of this variable at different seal clearances. A significant increase in axial tip deflection is produced by the use of the PR back plate. Figures 12 and 13 demonstrated the positive correlation between axial tip deflection and blow-down for a known seal geometry so a larger magnitude of bristle blow-down can be expected for the PR TEN-X seal.

Seal clearance is also shown in Fig. 20 to be more influential for the PR brush seal. While measured axial tip deflections increase significantly for larger seal clearances for the PR configuration, much smaller increases are experienced for the conventional seal. This is despite larger changes in seal clearance being considered for the conventional back plate.

The PR back plate configuration has therefore been shown to allow greater bristle blow-down than the conventional design through interrogation of key leakage pathways, the inter-bristle pressure field, and measurements of axial tip deflection. These increases in blow-down cause a decrease in leakage pathway $m˙3$⁠. Increased bristle blow-down was found for seal clearance C2 compared to the smaller C1, yielding only a small rise in leakage for the PR TEN-X seal. This contrasts significantly with the large degradation in sealing performance due to the corresponding clearance increase for the conventional TEN-X seal. A sudden change in seal clearance due to an off-design condition or rotor excursion will therefore be prevented from causing a substantial increase in leakage through a PR brush seal. This is because the bristles are able to follow the movement of the rotor surface and close much of the clearance area. Alongside other benefits such as reduced seal stiffness, this represents a significant advantage of PR back plate configurations in comparison to conventional designs.

This paper presents a novel application of DIC to track the individual tips of 152 bristles across all 12 rows of a linear, large-scale (TEN-X) brush seal. It is the first study to directly measure brush seal blow-down throughout the bristle pack and provides results for individual and row-averaged tip deflections.

Deflections in the radial plane, referred to as blow-down, were shown to vary with tangential position across the same bristle row. The advanced optical analysis employed in this study demonstrated the variation in the magnitudes of individual bristle tip displacement that results from small changes in geometry and inter-bristle frictional forces. However, this variation was shown to be sufficiently small allowing for row-averaged comparisons to be made. The results showed that the upstream bristle rows experienced greater blow-down relative to the downstream. This was attributed to the increased reaction and frictional forces, due to the proximity to the back plate. The analysis also highlighted the impact of bristle pack compaction during the initial pressure loading, which was less significant for the downstream bristle rows.

Bristle blow-down was shown to increase with seal clearance, with similar effects seen across the bristle pack. While an increase in seal clearance resulted in a rise in axial tip deflection for the upstream bristles, this effect weakened downstream and was negligible by the final row.

Interrogation of the inter-bristle pressure field of the TEN-X seal was conducted, demonstrating the different magnitudes of three key leakage pathways through the conventional and PR brush seals. It was highlighted that the PR seal increases blow-down, reducing leakage through the clearance area relative to the conventional configuration. Increased axial displacements for the PR back plate at corresponding operating conditions relative to the conventional design reinforced the increased blow-down experienced by the PR seal.

The increased blow-down with the PR back plate relative to the conventional design was shown to reduce leakage through brush seals with large clearances. The reduced stiffness of the bristles in the PR seal promoted blow-down which reduced the area underneath the tips and limited leakage in this region. Therefore, significantly smaller leakage flows are obtainable for PR brush seals operating with a clearance.

The authors wish to thank Andrew Langley at the University of Bath for his assistance with the maintenance and operation of the Large-Scale Test Facility. The authors also thank David Hollis at LaVision UK Ltd. for his support in the application and operation of the Digital Image Correlation apparatus and accompanying davis software.

• UK Engineering and Physical Sciences Research Council (Grant No. EP/P008232/1; Funder ID: 10.13039/501100000266).

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

A =

flow area (m²)

ATM =

atmospheric

B =

number of bristles in a given row

$ceff$ =

effective clearance (mm)

$cm$ =

seal clearance between the bristle tips and rotor surface (mm)

Con. =

conventional back plate

d =

bristle tip diameter (mm)

E =

bristle material Young's modulus (Pa)

$hf$ =

fence height (mm)

$hfb$ =

free bristle height (mm)

$m˙$ =

leakage flowrate (kg/s)

N =

packing density (mm⁻¹)

$Nb$ =

number of bristle rows

p =

static pressure (bar)

$pd$ =

downstream static pressure (bar)

$pu$ =

upstream static pressure (bar)

PR =

pressure relieving back plate

Q =

flow coefficient (K^1/2 s/m)

R =

gas constant (J/kg K)

T =

gas flow static temperature (K)

$tbp$ =

bristle pack thickness (mm)

TEN-X =

linear, large brush seal of ten-times scale

y =

radial location (mm)

z =

axial location (mm)

$zfb$ =

axial clearance between front and back plates (mm)

$zp$ =

pressure relieving back plate pocket depth (mm)

$δC$ =

calibration uncertainty

$δD$ =

total uncertainty for deflection result

$δy¯$ =

averaging uncertainty for row-averaged deflections

$Δp$ =

applied pressure load (bar)

$θ$ =

lay angle (deg)

$σ$ =

standard deviation

Normalized Parameters

$p*$ =

normalized static pressure

Y =

normalized radial location

Z =

normalized axial location

$ΦA$ =

normalized leakage flowrate

Measurements of radial deflection for each of the 13 bristles in Row 6 for the pressure relieving TEN-X seal at clearance C1 at $Δ$p = 4.4 bar are shown by the data points in Fig. 21. The mean value of these measurements is represented by the red dashed line, while the shaded area highlights the region within three standard deviations of the mean value. Statistical parameters of mean, range, variance, and standard deviation are quantified in Table 4. All measurements can be seen to fall within three standard deviations of the mean value, confirming the absence of outliers in the dataset.

This uncertainty analysis has been carried out according to the methodology of Moffat [23].

The uncertainty of the measured bristle deflection of each individual bristle was dependent only on the calibration error of the DIC system, detailed in Sec. 3.4. This calibration error was very small, corresponding to a focus area contained within 0.69 $μ$m $×$ 0.69 $μ$m. This is orders of magnitude smaller than the cross-sectional area of the bristle tip studied, which has a diameter of 1.02 mm, resulting in a small calibration uncertainty, $δ$C.

This process was followed to calculate the total uncertainty of the row-averaged radial deflection (blow-down) in Row 6 for the pressure relieving TEN-X seal with clearance C1 at $Δ$p = 4.4 bar. Row 6 lies in the middle of the bristle pack and the pressure load applied is large, so the uncertainty of this data point is representative of all row-averaged deflections presented in this study. The variation of radial deflection for each of the 13 bristles studied in this row at this pressure load is given in Fig. 21. The calculated uncertainties are stated in Table 5. $δ$$y¯$ is shown to be dominant, and $δ$C has negligible impact on $δ$D.