The future of aviation relies on the integration of airframe and propulsion systems to improve aerodynamic performance and efficiency of aircraft, bringing design challenges, such as the ingestion of nonuniform flows by turbofan engines. In this work, we describe the behavior of a complex distorted inflow in a full-scale engine rig. The distortion, as in engines on a hybrid wing body (HWB) type of aircraft, is generated by a 21-in diameter StreamVane, an array of vanes that produce prescribed secondary flow distributions. Data are acquired using stereoscopic particle image velocimetry (PIV) at three measurement planes along the inlet of the research engine (Reynolds number of 2.4 × 106). A vortex dynamics-based model, named StreamFlow, is used to predict the mean secondary flow development based on the experimental data. The mean velocity profiles show that, as flow develops axially, the vortex present in the profile migrates clockwise, opposite to the rotation of the fan, and toward the spinner of the engine. The turbulent stresses indicate that the center of the vortex meanders around a preferred location, which tightens as flow gets closer to the fan, yielding a smaller radius mean vortex near the fan. Signature features of the distortion device are observed in the velocity gradients, showing the wakes generated by the distortion screen vanes in the flow. The results obtained shed light onto the aerodynamics of swirling flows representative of distorted turbofan inlets, while further advancing the understanding of the complex vane technology presented herein for advanced ground testing.
A new trend in the pursuit for increasing efficiency and minimizing fuel consumption, emissions, and noise in modern aviation is the design and development of new configurations of aircraft that promote integration of airframe and propulsion systems. These aircraft concepts often rely on novel embedded engine configurations that involve fuselage boundary layer ingestion as a means of increasing efficiency [1–5].
Studies show that boundary layer ingestion can reduce aircraft noise, fuel consumption, and emissions. The reduction in noise comes from airframe shielding and lower jet velocities . The gains in efficiency come from less propulsion power due to the reduction in drag , wake recovery , and less kinetic energy wasted . These factors lead to lower fuel burn, reducing fuel consumption as much as 10% .
These novel configurations introduce potential problems in critical regimes of operation, as the ingestion of distorted flows, especially during take-off and landing, can result in adverse conditions for the fan rotor . It is imperative that these distortions are known and their impact fully understood in the initial steps of the aircraft design to ensure satisfactory operability. Total pressure and temperature distortions have been long analyzed [6–10], while swirl distortions, as relevant as they are, have only recently been successfully simulated and analyzed at laboratory and ground testing scales [11–13].
Figure 1 shows a rendering of NASA's hybrid wing body (HWB) aircraft concept, developed as part of the Environmentally Responsible Aviation program, with the main objective of reducing noise and fuel consumption. The experimental results report that this configuration can potentially reduce fuel consumption by more than 50% and improve noise conditions . The swirl angle profile in the inlet of a similar engine configuration as the HWB is also shown in Fig. 1. The profile was obtained from a CFD simulation and shows flow angle distortions ranging from −50 to 30 deg. The dangers of neglecting such deviations in flow angle include fan stall, performance alteration, and aeromechanical excitation [6,15,16], reinforcing the need to perform an analysis of the behavior of the inlet flow and the response of the engine to strong distortions.
The HWB distortion profile—indicative of airframe vortex ingestion—is the object of the swirl distortion study in this work. Here, the StreamVane™ technology [13,17] is used to generate a distortion profile inspired by the base distortion features of an extreme flight condition for an experimental design of a HWB engine inlet. Prior findings on StreamVane aerodynamics indicate that the dominant secondary flow dynamics are Reynolds number independent [17,18], and that the generation of the distortions is Mach number independent until vanes choke . For low-speed small scale wind tunnel experiments of single vortex distortions, secondary flow development is well-predicted by inviscid dynamics of axial vorticity . This makes the StreamVane an effective tool in generating distortions. However, there is still a need to describe the behavior of complex distorted flows in full-scale engine applications, especially near the fan.
For this work, stereoscopic particle image velocimetry (PIV) is used to measure three-component velocity fields of the generated flow at three different planes along the inlet of a full-scale research engine. The StreamFlow model is used to predict the secondary flow development based on the experimental data, in order to isolate spinner and fan effects. An analysis of the mean flow and turbulence produced by the distortion screen is presented in the results section to describe leading- and higher-order effects to the flow due to the StreamVane swirl distortion approach and how this flow develops as it approaches the engine fan.
The StreamVane method [13,17] is used to generate a predefined distortion in the inlet of a Pratt & Whitney Canada JT15D-1 research engine installed at the Virginia Tech Turbomachinery and Propulsion Research Laboratory. Particle image velocimetry is used to acquire three-component velocity fields at three different planes along the inlet of the engine for describing the development of the distorted flow as it approaches the fan in the turbofan engine. Details of the setup will be described in this section.
Computational fluid dynamics results of a simulated flow in the inlet of the engine of a HWB type of aircraft are used as the starting point for the design of the distortion screen. The secondary velocity profile at the inlet of the engine, presented in Fig. 2 and the swirl angle profile shown in Fig. 1, are used for determining the target profile to be generated by the StreamVane. This profile consists of a clockwise tightly wound vortex at approximately 180 deg and a bulk counterclockwise swirl. A computer-aided design model of a small-scale version of the screen is generated and printed using additive manufacturing. The small-scale 6-in version of the StreamVane is tested in a low-speed wind tunnel to ensure that the design is generating the desired profile. Then, after validation of the profile, it is scaled up to the full-scale inlet diameter, 21 in, and printed to be analyzed in the inlet of the research engine, resulting in the distortion screen shown in Fig. 2. A detailed description of the StreamVane method is presented by Hoopes  and Guimarães et al. .
The full-scale experimental setup consists of a bellmouth inlet, tunnel sections connected for inlet flow conditioning, a custom-built rotation stage to house and revolve the StreamVane, a special inlet duct section modified to allow optical access for the PIV cameras and laser sheet entry and exit, and the research engine. The flow is seeded from upstream of the bellmouth by two oil atomizers connected to blowers. Di-ethyl-hexyl-sebacate oil is used for seeding, yielding particles of an average diameter of 1 . The axial average bulk velocity of this flow is around 80 m/s (65% corrected fan speed), yielding a duct Reynolds number, , of 2.4 × 106, based on a duct diameter of 0.53 m and kinematic viscosity at 325 K of 1.80 × 10−5 m2/s.
A double-pulsed 532 nm Nd:YAG laser and two 4-Megapixel LaVision Imager Pro X 4M charge-coupled device (CCD) cameras are used to acquire stereoscopic PIV data at the predefined measurement planes, shown in Fig. 3. Details of the camera setup and laser properties are summarized in Tables 1 and 2. The CCD cameras are placed on an aluminum frame attached to brackets on a foundation isolated from the engine mount and built to minimize vibrations due to the operation of the engine, as can be seen in Fig. 4. The laser head is attached to the same frame, guaranteeing that the cameras and the laser will be subject to the same vibrations. A cylindrical lens connected to the exit of the laser head is used to expand the laser beam into a thin laser sheet. The laser sheet enters the PIV test section through an acrylic window located at the bottom of the test section and exits through another acrylic window at the top of the section, where measurement plane is labeled in Fig. 4. The PIV test section also has two 3-mm thick acrylic windows with antireflective coating to allow for optical access for the cameras, which are equipped with Nikon lenses and Scheimpflug adapters to correct the focus of the cameras at the measurement planes. The interior of the test section is painted flat black, and cameras and windows are covered with a black cloth during testing to reduce glare and laser reflection on the windows and tunnel walls. Testing is performed in the open cell after sunset to minimize light saturation on the second time exposure of image pairs.
The instrumentation is stationary and, due to limitations in the field of view of the windows, it is necessary to rotate the StreamVane to obtain a full sweep of the flow inside the 21-in diameter duct. The measurement volumes are represented by a diagram in Fig. 5. A thousand images are acquired with a 7 delay at 4 Hz for each position of the StreamVane, which is rotated 30 deg counterclockwise between measurements to allow for some overlap in order to improve statistics. The measurements are obtained and processed in LaVision DaVis 8.3.3, and the slices of data are further processed and stitched together through polar binning  in matlab.
The images are processed using stereo cross-correlation with background subtraction and time filter subtraction based on a Gaussian average of nine images. Multipass processing is used, with the first pass consisting of a square window of 64 × 64 pixels with a 50% overlap. Then, a square window of 32 × 32 pixels with 75% overlap is used for the following two passes. The vectors with a peak ratio (the ratio of the largest correlation peak to the second largest peak in an interrogation window) lower than 1.3 are deleted. A built-in median filter is used to remove and replace spurious vectors based upon neighboring vectors. Vectors that do not fall within an interval of the average 2 times the standard deviation are removed. Vectors that fall out of a range of 30 m/s in the secondary flow directions are deleted, and vectors that have an axial velocity of above 125 m/s are also deleted. The relationship between pixel size and flow space is also presented in Table 2. The spatial resolution corresponds to the size in flow space of the interrogation window in the last pass.
Results and Discussion
Flow in the inlet of a full-scale research engine is distorted by a 21-in diameter swirl distortion screen previously designed . The distorted flow, based on the distortion present in the turbofan inlet of a hybrid wing body airframe, is characterized experimentally. The StreamFlow model results are presented to describe the predicted evolution of the secondary flow profile, without engine interactions as core suction, spinner, and fan effects. The PIV results obtained are analyzed with two primary objectives in mind: to describe the development of the mean flow along the inlet at full-scale conditions, including interactions with the spinner, and to quantify the higher-order effects of the StreamVane to the flow, particularly the axial development of the vane wakes and turbulent stresses introduced into the flow by the distortion device. Near-wall data cannot be collected with PIV due to laser glare and limitations in processing window size, leaving a gap between the outer ring of the plots presented in this section, which represent the duct diameter, and the actual data.
To assess the development of the mean flow along the inlet, the mean secondary (in-plane) velocity profiles, normalized by the average bulk axial velocity are presented in Fig. 6. The profile used for designing the StreamVane is also presented for reference, but the 1.15D plane is not to be compared directly to the design for two reasons: the StreamVane method is supposed to generate the design profile right at the exit of the screen, which would be at 1.68D for this experiment, 0.53 diameters upstream of the first measurement plane, and the design profile has inconsistencies in the boundary conditions of some regions that cannot be replicated by the StreamVane, for example, at the lower right region.
The first feature of interest of this flow is the tightly wound vortex, which is very well delimited in all measurement planes. The bulk swirl affects the generation of the tightly wound vortex so that, at the 1.15D plane, it is pressed against the tunnel wall and broken into two smaller vortices, around 180 and 210 deg, as seen in the secondary velocity (Fig. 6) and in the streamwise vorticity profiles (Fig. 7). As the flow develops, the two vortices combine into a single vortex centered around 210 degrees at the 0.44D plane, which then moves clockwise due to its interaction with the wall which the bulk swirl to the 190-deg region in the 0.27D plane. The movement of the vortex can be seen in the constant radii vorticity plots presented in Fig. 8. This behavior is consistent with what is expected due to two-dimensional vortex dynamics, wherein the secondary flow development due to induced velocities from streamwise vorticity may be considered relatively independent from the axial flow, as seen in the StreamFlow model results shown in Fig. 9.
A quantitative examination of the flow velocity magnitudes indicates a loss in the strength of this vortex during the axial development, as well as a reduction in the vortex core diameter, indicating turbulence diffusion in the first case, explained further by the turbulent stresses of the flow. Axial stretching and stabilization from axial acceleration around the spinner explain the increase in streamwise vorticity magnitude for the 0.27D plane. While the 1.15D plane is dominated by the secondary flow, the 0.27D plane is where most of the influence of the spinner and the fan of the engine is observed. The magnitude of the in-plane velocity increases in the 0.27D plane in the region surrounding the spinner due to balancing of mass in the fan/core and the radial velocity increase caused by the spinner. This effect is isolated from the flow development in the delta profiles presented in Fig. 9, where the StreamFlow results (which do not include flow-spinner interactions), are subtracted from the experimental data obtained at each measurement plane downstream of the 1.15D plane. The regions with the highest delta values are the center of the profiles. For the 0.44D plane, that shows the effects of the core suction, and for the 0.27 plane, it is clear where the flow observes an increase in radial velocity, near the spinner. As mentioned, differences in the vortex region are due to vortex stretching and meandering.
This analysis can be broken down into two main regions of the flow: the near-wall, which is heavily influenced by interactions between the distortion screen and the growing boundary layer from the inlet duct walls, and regions away from the wall. As with the mean in-plane velocity profile, the effects of the distortion screen are more obvious in the 1.15D plane, since it is the first measured plane downstream of the StreamVane (at 0.536 diameters downstream of the device), where regions of axial velocity deficit are present due to the wakes introduced to the flow by the vanes. The 0.44D plot shows an increase of the axial velocity component, since the flow is being accelerated by the fan, and the vane wakes are not distinguishable from the flow anymore due to mixing. At the 0.27D plane, there is a general concentration of momentum toward the core. In contrast to the low momentum diffusive behaviors at 0–90 deg, high momentum regions appear to grow in regions of 240–270 deg and 90–150 deg.
A dynamic analysis of this flow is a next step in understanding the behavior of the tightly wound vortex. We approach this analysis presuming that the dominant effect of turbulent motions is turbulent diffusion of the secondary flow profile generated immediately downstream of the vanes. A number of factors influence the magnitude of the diffusion, including mixing from wakes of the turning vanes, shear-driven turbulence from velocity gradients in the swirl pattern, and azimuthal stretching due to the bulk swirl induced potential field of the rotor .
The normal components of the Reynolds stress tensor in the axial, , radial, , and azimuthal, , directions are presented in Figs. 11–13. Overall, the results exhibit turbulence intensities (single-component normalized rms velocities) of up to 15%—the same approximate magnitude as peak values in free shear flows . The axial Reynolds normal stress (Fig. 11) exhibits higher turbulence intensities in the same regions of interest mentioned previously: around the tightly wound vortex and near the wall. Around the vortex region, the 1.15D plane is subject to higher values of turbulent stresses, as expected, since it is closest to the exit plane of the StreamVane; thus, indicating that the secondary flow is highly unstable there. As the flow develops along the inlet, it is also accelerated, and the vortex is stabilized, leading to reduced turbulence intensities in the 180–210 deg region. The strength of the vortex is also reduced, indicating that the instabilities in the further upstream planes result in turbulent diffusion of the coherent axial vorticity prior to stabilization. The near wall region shows a growth in turbulent values from the 1.15D to the 0.44D planes due to boundary layer growth, but those values are reduced as the flow accelerates, confining the lowest momentum, highest gradient flow closer to the wall in the 0.27D plane.
The large-scale instabilities of the flow structure are also evident when analyzing the radial and azimuthal components of the Reynolds stress tensor, shown in Figs. 12 and 13. The higher values of the stresses around the vortex region in the 1.15D and 0.44D planes emphasize the fact that the vortex meanders around a preferred core region when influenced by the instabilities of the flow, and that it stabilizes as it approaches the fan face. The near-wall region also presents higher turbulence values, exhibiting inward radial turbulent diffusion of the thick boundary layer and shear layers from the near-wall vanes.
The StreamVane has been shown versatile and effective in generating complex distortion profiles, as the HWB profile presented in this work, single vortex distortions , and twin vortex distortions [19,26,27]. It is important, however, to acknowledge that it is a device added to the inlet of the engine, and its influence to the flow needs to be analyzed to ensure that possible adverse effects to the flow are minimized. Vane wake structures present in the 1.15D plane of the axial velocity profile, shown previously in Fig. 10, indicate that the wakes from the StreamVane vanes persist for at least 0.5 diameter downstream of the distortion device. This is also observed in the axial velocity gradients in the radial () and azimuthal directions (Figs. 14 and 15). As with the axial velocity profiles, the vane wakes clearly observed in the 1.15D plane mix and are not discernible anymore at the 0.27D plane. Furthermore, the near wall regions of the axial velocity gradient in the azimuthal direction (Fig. 15) indicate that the distortion screen is responsible for the formation of weak corner vortices, mainly at the 1.15D plane, for instance in the 30 deg region.
The Reynolds shear stress is examined in regards to understanding how the shear turbulence influences the development of the flow. In plane shear flows, the only terms that contribute to momentum transport are the axial/transverse shear. The physics of the effect can be conceptualized via the Boussinesq approximation for turbulent viscosity , indicating that the shear turbulence has a diffusive effect, directly correlated with the local velocity gradients. The Reynolds shear stresses, , presented in Figs. 16–18 exhibit structures that can be related with the ones of the gradients of the axial velocity component presented in Figs. 14 and 15. The 1.15D plane presents lower values for the shear stresses than the other planes, indicating that the turbulence is high in that plane, and that there is no significant coherent motion between the different velocity components. The shear stresses between the axial and each of the in-plane velocity components in the 0.44D plane (radial-axial, and azimuthal-axial) exhibit the spiral nature of the gradients of axial velocity in the inner regions of the flow, even though these stress values are low when comparing to the near wall region.
The shear stresses in the region of the tightly wound vortex, combined with the higher values of the normal Reynolds stresses observed in the same region, can be interpreted to indicate an increased diffusion of the vortex in the 1.15D and 0.44D, when compared to a better-stabilized or less-turbulent vortex, and to its behavior in the 0.27D plane.
A previously designed distortion device based on the StreamVane method to generate swirl distortion mimicking the behavior of a hybrid wing body type of aircraft is analyzed in the inlet of a full-scale P&WC JT15D-1 research engine. Stereoscopic particle image velocimetry measurements are taken at three different planes along the inlet of the engine: 1.15, 0.44, and 0.27 duct diameters upstream of the fan face of the engine. Mean velocity profiles, velocity gradients, and turbulent stresses at these locations are analyzed to describe the evolution of the flow along the inlet of the engine and the higher-order effects of the distortion screen to the flow.
There are two regions of interest in this flow, which are governed by different fluid dynamics flow components. The first is the tightly wound vortex, whose behavior can be described based on two-dimensional vortex dynamics, since it is not strongly affected by the axial flow present. As it approaches the engine, the vortex is axially stretched and stabilized by the accelerated flow. The near-wall region is also important and behaves based on boundary layer growth and interactions with the flow governed by the tunnel walls. The 1.15D measurement plane presents high turbulent stress values, indicating that the distortion device introduces large-scale instabilities to the flow. As the flow develops and approaches the fan, the effects of the presence of the spinner and the fan are observed, the flow is accelerated, and a more coherent motion is achieved.
Understanding the behavior of distorted inflows is crucial to the development of novel aircraft technology and more efficient, distortion-tolerant turbofan engines. Future steps in this work will involve the analysis of more fundamental flows, as single and twin vortices types of distortions, to better describe the physics involved in the distortion device.
Thanks to The Boeing Company (POCs: John Bonet and Ron Kawai), Pratt & Whitney (POC: Wes Lord) for the industry perspective on swirl distortion, and CAPES for financial support to Tamara Guimarães.
NIA and NASA Langley Research Center in association with NASA's Environmentally Responsible Aviation Project (NIA Cooperative Agreement No. NIA RD-2917). Project managers Fay Collier (LaRC), Hamilton Fernandez (LaRC), Greg Gatlin (LaRC), and Bo Walkley (NIA).
- D =
duct diameter, m
total-to-static pressure ratio
- R =
universal gas constant
- Re =
- T =
static temperature, K
radial velocity, m/s
axial velocity, m/s
azimuthal velocity, m/s
normal component of Reynolds stress tensor
shear component of Reynolds stress tensor
specific heat ratio
swirl angle, deg
kinematic viscosity, m2/s