High-resolution turbofan intake flow characterization by automated stereoscopic-PIV in an industrial wind tunnel environment

The experiments were conducted in the low-speed tunnel (LST) of the German-Dutch wind tunnels (DNW). The LST is an atmospheric closed-loop low-turbulence wind tunnel. It features a 9:1 contraction and a rectangular 3 × 2.25 m² test section where a maximum freestream velocity of 80 m s⁻¹ can be achieved. The intake test model used was representative of a modern UHBR configuration with a fan face hub to tip ratio (D_spinner /D) of 0.3 (see figure 1) achieved by the nose cone that was in place. The intake features droop (δ) and scarf angles (σ). Optical access for the laser sheet was provided through a transparent ring, separated by a slit to allow performing PIV measurements at two cross-flow planes upstream of the notional fan face plane separated by 0.04D (figure 1).

Zoom In Zoom Out Reset image size

The intake model was tailored made to facilitate the SPIV measurements. However, a range of additional, more conventional instrumentation was also included, to complement the SPIV measurements, which are reported below for completeness:

• An 8 × 10 rake configuration (8 rakes with 10 total pressure probes each) was used to measure the fan face, time-averaged total pressure distortion (red dots in figure 1-right). The rakes at 0°, 90°, 180° and 270° were further instrumented with a total temperature probe (thermocouple type). The rakes at 90° and 180° had two additional Kulites for dynamic total pressure measurement at these locations.
• 16 static pressure taps were used on the inner cowling behind the PIV slit to provide the circumferential static pressure distributions in the vicinity of the PIV plane (blue dots in figure 1 right). The ring was upstream of the fan face by 0.008D. Pressure taps were located every 22.5° with an offset of 10° from top dead center (blue dots in figure 1 right). In addition, a circumferential ring of 8 pressure taps was installed along the spinner, 0.008D upstream of the fan face location every 45°, with an offset of 10° (yellow dots in figure 1 right).
• To measure the lip static pressure distributions, 10 streamwise pressure sections were installed at 0°, 45°, 67.5°, 90°, 112.5°, 135°, 157.5°, 180°, 202.5°, 270° (Purple marks in figure 1 left). Each section had between 19 and 24 taps.

Two setups were investigated as part of the experiments at incidence and crosswind. For the incidence configuration, the model was mounted on a turntable (indicated in figure 2(a)) and was rotated about the PIV plane (referred to as α) from 0° to 35° for a set camera position. For the crosswind configuration, a ground board was placed at a representative vertical distance (h) of 0.30D from the lower lip highlight, while the model was rotated to 90° (refer to figure 2(b)). The flow direction is into the plane in figures 2(a) and (b) indicated by blue arrows. The model is rotated along the direction indicated in figure 2(a) (red arrow) for the incidence case.

Zoom In Zoom Out Reset image size

The required mass flow through the inlet was obtained via a system of blowers corresponding to a remote coupled engine/inlet configuration (Hodder 1981). The mass flow system (root blowers) was checked with a calibrated Bellmouth inlet which operates on the principle of Venturi flow meters (Smith 1985), and the errors in the computed mass flow were calculated to be less than 1% of the maximum designed mass flow rate (W_max).

For the PIV measurements, a multicamera optical setup with six LaVision Imager sCMOS cameras (2560 × 2160 pixels, 16 bits, and 6.5 µm pixel size) was used. Pre-tests conducted at the wind tunnel facility of TU Delft were used to de-risk the multicamera stereo-PIV approach. As part of these preliminary tests measurements were carried out employing stereo-PIV, tomographic-PIV and Lagrangian Particle Tracking using Shake-The-Box. Results from these experiments will be reported separately. For the measurements in the LST, a set of Nikon lenses with a focal length of 105 mm were employed. Each camera was fitted with an ISEL™ rotation stage and an automatic Scheimpflug to obtain a uniform focus quality and repeatability as the model configuration varied. This system enabled continuous data acquisition for multiple incidence angles and PIV planes without the requirement of manual adjustments. The cameras were placed outside the wind tunnel in two configurations—one for incidence (figure 3(a)) and crosswind conditions (figure 3(b)). Data for α corresponding to 25° to 35° were acquired in the incidence configuration.

Zoom In Zoom Out Reset image size

The required illumination was provided by two Quantel Evergreen Nd:YAG lasers, which were coupled and mounted on a traverse underneath the turntable (refer to figure 2(c)). The two laser beams, placed at 90° to each other (indicated in the illustration in figure 1), were used to minimise the effect of the shadow produced by the spinner. The automated traverse enabled the translation of the laser sheet from Planes 1 and 2 during data acquisition. The flow was seeded with DEHS particles of 1 µm median diameter. The pulse separation time was chosen to achieve an out-of-plane loss-of-correlation factor (F_o) larger than 0.75 (Keane and Adrian 1993). The experimental parameters are provided in table 1.

The cameras were calibrated for each measurement configuration (i.e., each α in the incidence case and the two PIV planes). The camera positions were chosen based on the prior experience in the pre-tests at TU Delft and the optimisation of the camera overlaps, which were analysed from CAD views of the test environment. The calibration was performed in three steps. Initially, the Scheimpflug axis was adjusted using the ISEL™ rotation stage, which was performed using the 2D dot pattern (figure 4(a)). In the second step, the appropriate Scheimpflug angle was obtained using the speckle image (figure 4(b)), where an optimisation algorithm determined the optimal focus quality from two regions on either side of the Scheimpflug axis, which was set in the earlier step. After the appropriate rotation angle and the Scheimpflug angle were achieved, the calibration images using the 3D calibration plate (refer to figure 4(c)) were acquired. The three plates were mounted at the location of the two PIV planes after removing the spinner. The parameters of each camera corresponding to the model orientation were saved prior to the measurements and could be recalled to minimise the time for image acquisition during the wind tunnel operation.

Zoom In Zoom Out Reset image size

Coverage of the domain of interest was quantified from the speckle pattern (figure 4(b)) at the different model configurations, i.e. at incidence (α = 0°, 15°, 25° and 35°) and crosswind. A quantitative indication of the overlaps achieved with the multicamera stereo-PIV system is provided in figure 5. The coverage of the domain of interest for α = 0°, 15° is partial as the position of the cameras (refer to figure 3(a)) was optimised for the incidence angle α = 30°. At least two cameras see 80% of the intake area at α = 0° and 93% at α = 15°. The cameras are set up in such a way that the regions imaged by less than two cameras at α = 0°, 15° and 25° (figures 5(a)–(c)) are regions where the flow is expected to be attached. The region of expected separation has good coverage with a minimum of 3 cameras. For the α = 25°, 35°, and crosswind configurations, at least 2 cameras see more than 95% of the intake area, thus enabling the velocity reconstruction in the complete domain (refer to figure 6). The curves in figure 6 indicate that the overall coverage of the inlet continues to increase as the angle of incidence increases from 15° to 35°. For the crosswind case, a complete coverage (∼98%) is observed, with a large area corresponding to 70% of the inlet area being viewed by all 6 cameras.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The incidence measurements were performed at a freestream Mach number of 0.2 for a range of angles of attack between 25°–35°, with an increment of 1°. To simulate different intake operating points, the model exit mass flow was set equal to $\widetilde {\dot m}_{{\text{corr}}} = \dot m_{{\text{corr}}}/\dot m_{{\text{corr}}}^{max}$ and 100%, where $\dot m_{{\text{corr}}}^{{\text{max}}}$ is the maximum corrected mass flow of the suction system. Under the wind tunnel conditions, the Reynolds number Re for the incidence cases, defined in terms of the intake inner diameter and freestream velocity, was approximately equal to 1.1 ·10⁶.

The crosswind measurements were performed primarily by varying the freestream velocity at a fixed value of mass flow through the intake, and secondarily by varying the intake mass flow at a fixed crosswind velocity. The crosswind velocity was varied in the range between 2 kts and 43 kts while the mass flow rate $\widetilde {\dot m}_{{\text{corr}}}$was varied between 30% and 102% of $\dot m_{{\text{corr}}}^{{\text{max}}}$. The intake performance in ground operation was investigated at a fixed representative vertical distance of h = 0.3D_hi (figure 1). The Reynolds number ($Re = {U_\infty }D/\nu$) for the crosswind cases varied between 0.3 × 10⁵ and 3.5 × 10⁵. To avoid early laminar separation, due to the low values of Re, boundary layer trips were used on the outer side of the intake surface. The complete test matrix of the experiments is provided in table 2, although only data for cases InW1, InW2 and XWS1 are addressed in the following sections of this paper.

Table 2. Test matrix for the incidence and crosswind configurations.

Incidence
Dataset	AoA (α°)	Free stream Mach number, M∞	Mass flow rate, ${\tilde{\dot{m}}_{\text{corr}}}$[%]
InW1	25°–35°	0.2	90
InW2	25°–35°	0.2	100
Crosswind
Dataset	Crosswind velocity, V∞ [kts]		Mass flow rate, ${\tilde{\dot m}_{\text{corr}}}$[%]
XWS1	15		34–102
XWS2	15		102–34
XVS1	4–43		86
XVS2	3.5–30		100

The PIV-data acquisition and processing was performed using the LaVision DaVis 11 software. The procedure was automated and interlinked with the wind tunnel operating software at DNW; this increased the productivity of the measurements and reduced the wind-on time for data acquisition. The measurements corresponding to the incidence sweep, which consisted of 60 data points (section 2.3) were completed in about 4 h.

Each data point measurement consisted of a sequence of 1000 double-frame images. Wind tunnel vibrations occurred during operation, which affected the camera images. To correct this, markers were placed on the outer surface of the inlet to facilitate the application of a shift and vibration correction algorithm (dots on the outer surface of the inlet in figure 7(a)). Image pre-processing was performed to remove the background light by normalising the pixel intensity with respect to the time-averaged value. After that, a minimum intensity over the subsequent 15 frames was subtracted from the images (figure 7(b)). In addition, the flow conditions in crosswind resulted in condensation of water around the ground vortex, which saturated the pixels in that region and inhibited the stereo reconstruction. To prevent spurious vectors, an algorithmic mask based on a threshold of pixel intensity was applied over this portion of the vortex. After this, the region of the image containing the particle images was extracted by the application of a geometric mask (figure 7(c)).

Zoom In Zoom Out Reset image size

The image interrogation was performed using multi-pass cross-correlation with window refinement (Scarano and Reithmuller 2000). In the final iteration, windows of 48 × 48 px with Gaussian weighting were used with an overlap factor of 75%. The resultant vector spacing was 1.5 mm (corresponding to 0.006D, 0.6 vector mm⁻¹), which resulted in 18 750 velocity vectors across the region of interest. To quantify the range of resolvable velocity scales, the dynamic velocity range (DVR, Adrian 1997) is determined as the ratio of the maximum measured velocity in the inlet plane and the standard deviation of the velocity distribution in the separated case. The details of the image processing parameters and estimates of the measurement dynamic range are summarised in table 3.

The average uncertainties in the particle displacement were calculated based on the correlation statistics described by Wieneke (2015). Figures 8(a) and (b) shows the uncertainties in mean velocity magnitude scaled by the mean axial velocity through the inlet. Figure 8(a) corresponds to the inlet at incidence where the flow is separated, and figure 8(b) to the crosswind case with the ground vortex and flow separation. The uncertainties in velocity are less than 2% in both configurations.

Zoom In Zoom Out Reset image size

The steady pressure measurements at the AIP were acquired over a period of 20 sec at an acquisition frequency of 625 Hz. The data were averaged over the duration of the PIV acquisition (for 1000 images, approximately 2 min).

For the current analysis, the PIV derived velocity maps are presented within the region between 0.3 R < r < 0.97 R, to ensure that no invalid measured data influence the postprocessing. Flow separation inside the intake was identified by means of the isoline of zero axial velocity for each snapshot, and the coordinates of the approximated centre of the separated region were calculated. The separation behaviour was further analysed considering the isentropic Mach number derived from the static pressure measurements around the intake wall:

$$\begin{equation}{M_{is}} = \sqrt {\frac{2}{{\gamma - 1\,}}\left[ {{{\left( {\frac{{{P_{0,\infty }}}}{p}} \right)}^{\frac{{\gamma - 1}}{\gamma }}} - 1} \right]} .\end{equation} \tag{ 1 }$$

For the crosswind cases where velocity measurements in the vortex core were available, the vortex core position at the PIV plane was identified for each instantaneous flow field by detecting the maximum value of the Q-criterion (Hunt et al 1988), defined as:

$$\begin{equation}Q = \frac{1}{2}\left(\|\Omega^2\| - \|S^2\| \right)\end{equation} \tag{ 2 }$$

where ${{\Omega }}$ and $S$ are the vorticity tensor and the strain rate tensor, respectively. Based on the definition, positive values of Q (i.e., Q > 0) are indicative of areas in the flow field where a vortex is present, provided that local static pressure is lower than the free-stream value. To avoid the interference of the vortical structures located inside the separated region, when present, and of the boundary layer region, only data within the region 135°—240° and r < 0.96 R was considered when performing the vortex core tracking.

The contours of the time-averaged axial velocity, non-dimensionalised by the time-average area-average axial intake velocity, are reported in figure 9 and in figure 10, for the case at ${\widetilde {\dot m}_{\text{corr}}}$ = 90% and ${\widetilde {\dot m}_{\text{corr}}}$= 100%, respectively (cases InW1 and InW2 in table 2). The velocity fields correspond to a freestream Mach number of 0.2 and an incidence angle in the range 25°–35° for both intake operating conditions. The separation within the intake can be identified by the presence of a region of reversed flow, with negative axial velocity (figure 9(d)). From the observation of the flow fields reported in figures 9 and 10, starting from an attached flow condition (figures 9(a) and (a)), as the incidence angle increases, the flow inside the intake separates from the intake surface lip (figures 9(d) and 10(c)), generating a region of detached flow that increases in size with the angle of attack. Similar flow behaviour and flow topology, for which an intake exhibits lip separation due to high incidence angle, has been previously reported in several experimental works (Hodder 1981, Larkin and Schweiger 1992), which however relied only on conventional distortion measurement techniques, i.e. total pressure rakes.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The nature of this separation can be assessed by considering the peak isentropic Mach number variations with the angle of attack, as reported in figure 11, for the two intake mass flow settings, respectively. The isentropic Mach number (equation (1)) was calculated from the static pressure measurements around the intake surface at four different azimuthal coordinates on the bottom half of the intake surface. The angle at which the separation occurs for 90% ${\widetilde {\dot m}_{\text{corr}}}$ is 1° lower than for the case with 100% ${\widetilde {\dot m}_{\text{corr}}}$.

Zoom In Zoom Out Reset image size

As the angle of attack increases, the flow accelerates around the intake lip, resulting in a highly supersonic isentropic Mach number. At relatively low values of the incidence angle before the separation occurs, the flow is still attached to the intake surface, even though the boundary layer thickness becomes greater (figure 10(c)), indicating incipient flow separation. At the separation angle of attack, a shock is formed over the intake lip at 180° and 202.5°, causing the isentropic Mach number to drop to subsonic values and the flow separation.

The normalised axial velocity at the measurement plane with increasing mass flow rate is shown in figure 12 for a crosswind velocity of 15 kts and ${\widetilde {\dot m}_{\text{corr}}}$ between 34% to 102% (case XWS1 in table 2). As shown for the cases at high incidence, the separation within the intake can be identified by the presence of a region of reverse flow, with negative axial velocity. As the mass flow through the intake is increased, the intake flow goes from an initial separated condition (${\widetilde {\dot m}_{\text{corr}}}$ = 34%) to an attached condition (${\widetilde {\dot m}_{\text{corr}}}$ = 65%) and then separates again (${\widetilde {\dot m}_{\text{corr}}}$ = 94%). Due to the non-axisymmetric geometric features of the intake (scarf and droop angles) and the presence of the ground plane, the separated region is forced towards the top-half of the plane on the side exposed to the cross flow. As the corrected mass flow rate increases from 34% to 56%, the size of the distorted region decreases until the flow reattaches, at ${\widetilde {\dot m}_{\text{corr}}}$ = 65%. As the ${\widetilde {\dot m}_{\text{corr}}}$ is further increased from 85% to 94%, the flow separates again.

Zoom In Zoom Out Reset image size

Similar flow behaviour for an intake in crosswind, for which two separation limits can be identified, has been previously reported in several experimental (Quemard et al 1996, Lecordix et al 1996) and computational studies (Colin et al 2007, Zhang 2021). For low values of the corrected mass flow ${\widetilde {\dot m}_{\text{corr}}}$ up to 56% of the maximum (figures 12(a)–(c), respectively) the separation is diffusion-driven. In these cases, the intake exhibits a wide area of reversed flow on the windward side and the distributions of the isentropic Mach number shown in figure 13, which provide information about the peak values, the position, and the extent of the separation, present a subsonic peak value. In these cases, there is a strong flow deceleration around the intake lip, highlighted by the presence of a plateau in the isentropic Mach number profiles in figure 13, causing the boundary layer to separate and form a large area of reversed flow. At high ${\widetilde {\dot m}_{\text{corr}}}$ (from ${\widetilde {\dot m}_{\text{corr}}}$ = 94% to ${\widetilde {\dot m}_{\text{corr}}}$= 102%, figures 12(f) and (g)), the separation is shock-induced, since the maximum isentropic Mach number for these cases, as seen in figure 13, is supersonic. The flow is strongly accelerated around the lip until it becomes locally supersonic. This generates a shockwave strong enough to cause the separation of the boundary layer. The presence of the shock wave can be identified in figure 13 by the presence of an abrupt drop in the isentropic Mach number. Between the two separation limits (${\widetilde {\dot m}_{\text{corr}}}$ = 65% and ${\widetilde {\dot m}_{\text{corr}}}$ = 85%, figures 12(d) and (e)), the flow remains fully or mostly attached over the lip and the diffuser.

Zoom In Zoom Out Reset image size

Intake flows in crosswind are known to be characterised by a high level of unsteadiness due to the complex nature of the three-dimensional flow and the coexistence of flow separation and ingested vortex (Murphy and MacManus 2010, Wang and Gursul 2012). The instantaneous and time-averaged vortex core locations for the case at 15 kts with increasing (case XWS1 in table 2) are reported in figure 14, along with the probability of the vortex to occupy a specific location on the measurement plane.

Zoom In Zoom Out Reset image size

It can be observed that as the corrected mass flow is increased from 34% (figure 14(a)) to 65% (figure 14(d)) the ingested vortex is oscillating within a relatively narrower area around the time-average location, suggesting a decreasing level of unsteadiness. Moreover, at ${\widetilde {\dot m}_{\text{corr}}}$ = 34% (figure 14(a)), the area around which the vortex core location is oscillating exhibits an elongated elliptical shape compared to the one at ${\widetilde {\dot m}_{\text{corr}}}$ = 65% (figure 14(d)), which presents a nearly circular shape. The difference in shape suggests that the vortex oscillation at ${\widetilde {\dot m}_{\text{corr}}}$ = 34% is predominantly in the direction parallel to the crosswind velocity. Conversely, at ${\widetilde {\dot m}_{\text{corr}}}$ = 65% the vortex oscillates in both directions, parallel and perpendicular to the cross flow, affecting a larger radial portion of the plane.

In terms of the lip separation, the instantaneous and time-averaged position of the centre of the separated area for three cases at 15 kts across a range of corrected mass flows ${\widetilde {\dot m}_{\text{corr}}}$ is reported in figure 15. The case with ${\widetilde {\dot m}_{\text{corr}}}$ = 34% was found to be dominated by a diffusion driven separation, the case with ${\widetilde {\dot m}_{\text{corr}}}$ = 65% was mostly attached and the case with ${\widetilde {\dot m}_{\text{corr}}}$= 102% showed a shock-induced separation. The centre of the separated region was defined as the centre of the circular sector enclosing the portion of the measurement plane where the reversed flow was present.

Zoom In Zoom Out Reset image size

Based on the instantaneous positions of the centre of the separated area (figure 15), identified as the centre of the region with negative axial velocity, it can be observed that for the case with diffusion-driven separation (${\widetilde {\dot m}_{\text{corr}}}$ = 34%, figure 15(a)), the radial extension of the separation is higher with respect to the cases characterised by shock-induced separation (${\widetilde {\dot m}_{\text{corr}}}$ =102%, figure 15(c)). For the case where diffusion-driven separation is present, the radial extent of the separated area covers almost 25% of the plane radius (figure 15(a)). On the other hand, for the case characterised by shock-induced separation (figure 15(c)), the separation is confined inside the outermost part of the measurement plane, near the outer surface of the intake, extending across a smaller part of the plane. It can also be noted that in the case at ${\widetilde {\dot m}_{\text{corr}}}$ = 65% the time-average flow field is attached (figure 12(d)).