Experimental and Numerical Comparison of Two Turbine Vane Frame Designs

The current development of state-of-the-art civil aircraft engines tends toward a substantial increase in the bypass ratio. Evidence of this can be seen in current architectures such as the geared turbofan [1] and the upcoming open rotor concept announced by CFM [2]. A fast-spinning low-pressure shaft is a common feature in both concepts. One of the distinctive differences of fast-spinning low-pressure turbines (LPTs) is the diameter reduction, which further impacts the intermediate turbine duct (ITD) located upstream of it. For conventional ITDs, the main aerodynamic purpose is to overcome the typical radial offset of the high-pressure turbine (HPT) with respect to the LPT within a minimal axial distance due to weight considerations. Besides the aerodynamic function, the ITD has also the structural purpose of housing the high-pressure shaft rear bearing. Access to critical bearing support lines is provided by airfoil-shaped fairings, or struts, incorporated in the duct. The reduced radial offset caused by the fast-spinning LPT allows to combine the task of the LP vanes with the typically non-lifting struts. The result is an S-shaped duct with turning struts, typically known as the turbine vane frame (TVF).

The work of Göttlich [3] shows that the aerodynamic behavior of such S-shaped ducts is highly affected by their two bends which cause the formation of radial pressure gradients through the duct. Spataro et al. [4] described two additional pressure gradients imposed by the turning struts; a circumferential pressure gradient from the pressure to the suction side and a radial pressure gradient at the outlet due to the presence of swirl. Clark et al. [5] demonstrated that the introduction of splitters serves to partition secondary flow vortices into smaller fractions of the overall streamwise circulation. This results in reduced secondary kinetic energy at the outlet when considering a splitter count of two instead of one. Spataro et al. [6] found that the integration of splitters improved the outlet flow conditions, and hence the LPT stage efficiency, when compared with a no-splitter design. Investigations at the Van Karman Institute [7,8], where three splitters were added to an existing low aspect ratio vane row, led to the conclusion of incidence sensitivity due to the potential field of the strut. In contrast, Pramstrahler et al. [9] determined an incidence insensitivity between −20 deg and +20 deg for an investigated two-splitter TVF.

In this paper, both investigated geometries feature the same amount of passages and both include splitters—TVF1 features one splitter, while TVF2 has two splitters per strut passage. The focus of this work is on the impact of various inlet flow conditions on the flow field and performance of the aforementioned TVF designs. Inflow variations are achieved by different purge flowrates (PFRs) or incidence changes. The TVF performance is assessed based on the pressure loss and important outlet flow conditions, such as flow angles (yaw and pitch) and Mach number variation.

This work mainly presents experimental results obtained in the transonic test turbine facility (TTTF). Furthermore, computational fluid dynamics (CFD) results at the aerodynamic design point (ADP) with nominal purge flowrates are included in order to provide additional insights into the observed flow structures.

All tests were performed at the TTTF at Graz University of Technology, Graz, Austria. The rig consists of two separately mounted shafts (HPT and LPT) with mainstream flow primarily supplied by a 3 MW main compressor station (CS1) arranged in an open circuit, as depicted in Fig. 1. Both rotors are overhang mounted to simplify modifications. Furthermore, the entire frame of the LPT can be moved axially. This results in flexible axial design options and simplified access to the rotors in the event of assembly and disassembly. The HPT shaft, which cannot be moved axially, drives a three-stage radial brake compressor (BC), which provides additional mass flow for the test rig. As a result, the TTTF can provide up to 22 kg/s of mainstream flow at 4 bar inlet pressure. The power extracted by the LPT is dissipated in a 700 kW water brake. The following 560 kW suction blower extends the achievable overall pressure ratio. In addition, fluctuations of the ambient pressure do not affect the rig exit pressure. More details on the HPT side can be found in Erhard and Gehrer [10]. An in-depth description of the operating capabilities of the rig is provided by Neumayer et al. [11]. Detailed information on the LPT spool design can be found in Hubinka et al. [12].

To enable purge flow supply to the hub and shroud cavities of the turbines, a secondary air system was introduced in 2015, fed by an independent 1.1 MW compressor station (CS2). To provide the desired purge flow temperature to each of the cavities, air is mixed from a cooled and uncooled reservoir. Each of the main purge supply lines splits into multiple sublines to ensure uniform distribution of the purge air around the circumference. More information can be found in Steiner et al. [13].

The presented test vehicles feature the same HPT stage followed by a differing TVF and LPT rotor. To characterize the investigated TVFs, the duct inclination and the streamwise area ratio (AR) are used, as shown in Fig. 2.

The streamwise area ratio is calculated using the area perpendicular to the rotary axis of the TVF inlet plane (plane B) and the outlet plane (plane C), projected in the streamwise direction using the mass flow averaged flow angles α and γ obtained by the five-hole probe (FHP) measurements. By consideration of the inlet flow angles, TVFs with non-axial inflow can be characterized. Values less than 1 indicate an accelerating ITD and values larger than 1 a diffusing ITD. An example of a diffusing ITD recently investigated at TU Graz [17] is added amongst the two considered TVFs in Fig. 2 to emphasize the capabilities of the streamwise area ratio as a means of characterizing all types of ITDs.

The two TVF geometries considered in this work are in the same order with respect to the duct inclination. However, the AR_sw shows a distinctive difference, resulting in a more pronounced acceleration for TVF2 (AR_sw < 0.65). The strut count is the same, while TVF2 has two splitters per passage instead of the single-splitter design of TVF1.

Both setups have four measurement planes shown in Fig. 3: plane A upstream of the HP vanes, plane B downstream of the HP-rotor, plane C at the outlet of the TVF, and plane D downstream of the LP-rotor. The distances with respect to the axial chord length are reported for planes B and C in Table 1.

The operating conditions of the TTTF are primarily quantified and adjusted by the speed of the rotors (±0.1% uncertainty), the mixing chamber temperature (±0.3 K uncertainty), and the rotors' total-to-total pressure ratio (±0.5% uncertainty).

The operating point of the HPT, and hence the inflow conditions to the TVF are similar, allowing for a fair comparison of the two TVF designs showing the operating point of the test rig normalized by the TVF1 conditions. The off-design operating points presented in this paper are achieved by varying the HPT cavities (H-TF, H-TA, H-HF, and H-HA) PFR, while keeping the other quantities (rotor speed, rotor pressure ratio, and mixing chamber temperature) constant or by varying the incidence, via HP-rotor speed variations. The 100% PFR is defined for all purged cavities as a fraction of the total inlet massflow and is measured by two venturi flowmeters (2%). However, the massflow of every purge supply line is measured with McCrometer’s V-cone flowmeter (0.5%). Among the nominal PFR (100%), there are three derivatives: the no purge case, where all supply lines associated with the HPT are closed, the low purge case (50% PFR), and the high purge case (200% PFR). The purge massflow distribution of the HPT is shown in Table 2, normalized by the AFT hub mass flow. The fifth purge supply line that provides mass flow for the Hub FWD LPT (L-HF) cavity was set at 100% PFR rate for all operating points.

The swirl derivatives contain a +4 deg incidence for TVF2 and a +6 deg for TVF1 (Fig. 3; +α).

All presented numerical results are based on a 3D steady Reynolds-averaged Navier–Stokes simulation utilizing the commercial code ansys cfxrv2020 r1. The domain is split up into two subdomains: the stationary TVF from plane B to slightly downstream of plane C, and the rotating LPT ending at plane D, connected via a mixing plane. The inlet 2D-boundary conditions are based on the five-hole probe measurements except for the turbulent intensity, which relies on hot-wire measurements. The applied endwall assumptions for all quantities are described by Krajnc et al. [18]. Further, the SST k–ω turbulence model was utilized [19]. At the outlet, the mass flow averaged radial distribution of the static pressure obtained with the FHP acts as a boundary. Both hexahedra meshes were created in the context of a Master’s thesis, including a sensitivity analysis [20,21]. Furthermore, the meshes of the duct fulfill the y⁺ criterion for low Reynolds modeling, where the mean and max. values are shown in Table 3.

CFD validation is achieved by comparison of the total pressure in the crucial plane C, as shown in Fig. 4. The nominal-purge condition of both cases shows very good agreement with the measurements. Moreover, the no purge case of TVF2 matches the measurement and captures the same trends (similarity between hub and ∼8% span, offset ∼80% span).

The entire data set of both test vehicles is obtained with the same five-hole probe, whereby raw pressures are acquired with the same pressure transducer for a fair back-to-back comparison. The measurement grid covers one TVF pitch, a field of approximately 20 × 50 points (radial × circumferential). The probe head, which has an inclination of 105 deg contains a shielded and calibrated type K thermocouple for total and static temperature measurements, respectively. The probe is manufactured and calibrated by the Institute of Jet Propulsion and Turbomachinery of RWTH Aachen University. Raw pressure data are acquired with a 15 PSI NetScanner 9116 pressure scanner (±52 Pa).

Further, raw temperature values are red by EX10xxA series temperature input module (±0.2 K; for a measurement range of 0–100 °C). The measurement for each point takes 6 s with a sampling frequency of 5 Hz. The collected data of 5 × 30 pressure readings and 30 temperature readings are averaged and further used to calculate the flow quantities with the calibration polynomes in the post-processing stage. The five-hole probe uncertainties are reported in Table 4.

The flattened Pitot tube was used for measurements in plane C of TVF2 only to acquire data as close as possible to the hub. Given the FHP dimensions and the steep inner endwall in the respective measurement plane, the lowest achievable FHP position (∼8% span) would have been insufficient to resolve the near hub region. However, with the flat Pitot tube, measurements were obtained starting at a minimal endwall distance of ∼0.5% span up to midspan. The probe was pre-yawed based on the FHP results initializing every point independently. For the measurement points between the innermost FHP reading and the hub, a linear interpolation of the FHP reading and the averaged yaw angle at the hub obtained by CFD simulations was utilized. Prior to the measurements, the sensitivity behavior was checked for the uncalibrated probe, showing insensitive characteristics of 1% for ±15 deg. This data set was acquired in the measurement plane while the rig operated with test conditions.

For the oil flow visualization, a mixture of motor oil and titanium dioxide was adopted. The premixed paint is injected via a piston into desired wallstatics while the rig is running at the operating point. This approach is mainly applied for the injection into purge cavities in order to trace back cavity flows.

For the case of the total-to-total pressure loss (Eq. (3)) Y = f(p_t,B, p_t,C) with the uncertainties reported in Table 3. In general, all shown uncertainty bars are normalized with the corresponding factor.

The first part of the results focuses on the outlet flow of the HP rotor in order to identify the flow structures that affect the TVF aerodynamics. Second, the distinctive flow features are then discussed for the two examined cases and work out their performance differences in terms of pressure loss for different PFR and incidence, respectively. Ultimately, the focus is on the outlet flow condition and its off-design behavior.

All contour plots shown in this paper are oriented ALF. To improve the quality of the plots, the measurement results have been interpolated to a finer grid.

Figure 5 shows the normalized total pressure in plane B for 100% PFR obtained during the measurement campaign of TVF2, which is representative of all investigated operating points for the two cases. Between the hub and ∼40% span HP-vane related structures can be seen: low p_t-zones 1–3 V can be assigned to the LPV_V, which modulates the high p_t band, forming three distinct structures 1R–3R. In the upper span region (60%–85%) the same behavior is noticeable caused by the upper passage vortex (UPV) of the HP vanes (4–6 V and 4R–6R). Further, the potential effect of the strut (leading edge indicated with a dashed line) is observable, resulting in slightly higher total pressure, which suppresses the appearance of the wake (dashed circle).

Figure 6 shows the circumferential mass flow averaged and corrected total pressure profile in plane B for both investigated setups. Figure 6(a) depicts the baseline case (PFR 100%) for TVF1 and TVF2 normalized with the TVF1 mean value. In general, one can observe a very good overall agreement between the two cases, leading to the conclusion of a fair comparison. The biggest variation appears at ∼95% span, and is due to the different shroud contours between the two test vehicles, which affects the static pressure gradient at the exit of the HPT, resulting in a more distinctive tip leakage vortex (TLV).

Figures 6(b) and 6(c) present the impact of varying the PFR (0%, 50%, and 200%) normalized with its mean value. For both cases, one can notice that secondary flow structures migrate toward midspan as observed by Zerobin [23]. Furthermore, increasing the PFR suppresses the tip leakage vortex as it can be most noticeably seen in the case of TVF1. While there is a distinctive peak for no purge, the local maxima weakens by injecting more purge flow until 200% PFR, where the presence of the TLV is not detectable anymore. The same but not that pronounced behavior can be seen in Fig. 6(b) representing TVF2. The positive incidence case for both setups is not shown in this graph due to its indistinguishable shape compared to 100% PFR.

The analysis of outlet flow conditions relies on total pressure and pitch angle measurements, as illustrated in Fig. 7. Figure 7(a) shows the normalized total pressure at the outlet of TVF1. The corresponding velocity vectors are also included in the figure to discern fluid migration patterns. The TVF exit flow is characterized by a number of large low-pressure regions as a result of the presence of various secondary flow structures.

The most prominent low-pressure region is situated next to the suction side of the strut at the lower half of the span. The large spatial extent of this region in the circumferential direction is due to the presence of two vortices: the PS-HSV of the splitter and the LPV formed between the splitter pressure and strut suction sides. The large radial extent of this region, on the other hand, is due to the radial migration of the UPV (core marked with A in Fig 7(a)) and the tip leakage vortex from the upstream HPT_R (see Sec. 3.1). The formation and migration of the aforementioned vortices is strongly dependent on certain operating conditions and design characteristics of TVF1 such as the positive incidence at the inlet and the low aspect ratio airfoils which promote the early formation and enhancement of the horse-shoe and passage vortices, and the duct bends which impose strong radial gradients and associated radial migrations of all secondary flow features propagating through the duct.

Further, two distinct low-pressure regions can be seen adjacent to the suction side of the splitter. The low-pressure region nearest to the hub is due to a combination of the LPV formed between the strut pressure and splitter suction sides and a shed vortex formed due to a radial velocity gradient at the trailing edge in that region (see discussion of Fig. 8). The radially stretched low-pressure region between 40% and 80% span is due to a combination of the UPV in PS-passage and the splitter upper SS-HSV. The duct design characteristics are affecting the development and migration of also these features.

Figure 7(b) shows a contour plot of the pitch angle at the exit of TVF1. Adjacent to the splitter suction side, the vortices induce radial fluid migration toward the midspan, as indicated by the positive and negative pitch angles at high and low spans respectively (DW_Sp and UW_Sp). Next to the suction side of the strut however, a large negative pitch angle region shows that the flow pitches down across almost the entire span, indicating a pronounced downwash. The concentration of the various vortical structures at the lower half of the span next to the strut suction side discussed previously is due to this strong downwash (DW_St).

The effect of the secondary flows and their migration on the airfoil surface-flow is shown in Fig. 8 as obtained from CFD. The pressure side of the struts and splitters is unaffected by any of the secondary flow features that develop through the duct and is hence predominantly characterized by 2D flow. Streamline curvature effects due to the duct's first and second bends are visible on this surface, especially in the case of the strut (marked with red dashed lines in the figure). The suction side of the TVF airfoils near the trailing edge is, as previously alluded, heavily affected by the secondary flow vortices that propagate through the duct. The streamlines in this region clearly show the influence of these structures for both the strut and the splitter. In the case of the strut, and in agreement with the results in Fig. 7(b), the downwash region is shown to influence the majority of the span of this airfoil. A small upwash region is also shown but this is confined within the hub fillet; an area not identified in Fig. 7(b) as is outside of the measurement range of the 5HP. In the case of the splitter, a more pronounced upwash can be seen but the downwash region is still the largest of the two. The significance of the upwash region on the splitter suction side is that along with a mild downwash on its suction side results in the generation of a shed vortex that contributes to the low total pressure region near the hub identified in Fig. 7(a). In general, the radially inwards and outwards migrations on the suction side of the TVF airfoils are expected to “spoil” their aerodynamic performance with implications on the loading distribution in the affected part of the airfoil.

Figure 9(a) shows a contour plot of the normalized total pressure loss at the outlet of TVF1 for the 0% PFR case defined by Eq. (3). The pressure loss is divided into three main regions: L1, L2, and L3. The pronounced loss region L1 adjacent to the SS of the strut is due to the various secondary flow structures that accumulate in that region and their interactions. The low loss region above L1, is associated with the downwash effect and appears with a higher pressure loss compared to the remaining part of the passage but not in the same order of magnitude as L1.

The other two loss regions can be found adjacent to the splitter suctions side. The radially stretched region L2 is due to the UPV of the PS-passage. Region L3, near the hub is as a result of the LPV of the PS-passage and the “shed” vortex due to the radial velocity gradients on either surface of the splitter in that region.

Figure 9(b) shows differential plots of the total pressure at the exit of TVF between the 0% PFR case (reference) and the various purge flow derivative cases, as defined by Eq. (4), to highlight the impact of the purge flowrate on the TVF loss. The lower limit of the color bar in this figure is set to zero as no negative values (higher P_t in the case of derivatives compared to the reference case) were observed. Red zones indicate regions where the total pressure of the purge derivative is lower than the reference case, thus signifying additional pressure loss. To improve the interpretation of the results, the iso-lines of the pressure loss features (L1, L2, and L3) of the reference case from Fig. 9(a) are embedded in all diff-plots.

The differential plot for 50% PFR shows a growth of the L1 loss core on three positions, D1, D2, and D3. A careful look at the contour line of the L1 loss core from the 0% PFR case indicates that this loss core has indeed expanded and not strengthened as a light blue color indicating no total pressure change can be seen at the center of the reference sub loss cores (marked with X). The expansion in loss in these positions is due to the thickening of the inlet boundary layer at hub and case which results in an earlier formation and strengthening of the various secondary flow features that affect that region of the TVF exit.

The increase in pressure loss in position D1 can be assigned to the downwash structure of the struts SS that contains most of the inlet flow above ∼90% span. As discussed in Sec. 3, the p_t-profile in plane B for TVF1 shows a weakened TLV by increasing the PFR. Subsequently, the high-momentum fluid appears at smaller radii, which enables the formation of a stronger UPV assuming an unaffected circumferential pressure gradient. The differences at D2 and D3 are caused by the growth of a shed vortex, which implies that the flow angles at the trailing edge have slightly changed.

Increasing the PFR to 100% results in the same loss expansions as in the 50% PFR case. In this case however, the loss positions, D1–3, appear to be enhanced indicating an increase in the loss. The reason for this is the strengthening of the secondary flow vortices affecting this region as the endwall flow at the inlet of the TVF is further weakened by the introduction of larger purge flow amounts.

Further increasing the PFR to 200% results in an even larger expansion and strengthening of the loss cores with positions D1–3 growing in both radial and circumferential directions. The amount of purge supplied is so high that an impact is also seen on the loss generated by the secondary flow features adjacent to the splitter suction side.

The description of the outlet flow of TVF2 follows the same approach as for TVF1. Figure 10(a) depicts the total pressure in plane C overlapped with the corresponding flow velocity vectors.

In general, the flow field at the exit of TVF2 appears to be more “conventional” than the one of TVF1 with the low-pressure lobes as a result of the upper and lower passage vortices situated closer to the respective endwalls.

The reason for this more “conventional” secondary flow distribution in the case of TVF2 as opposed to TVF1 can be better explained by considering the static pressure gradients at the front part of the duct, as shown in Fig. 11. The ∼45 deg contour lines of the pressure field in B_{x, TVF1}, suggest that the pressure gradient imposed by the first bend (radial) and the circumferential pressure gradient are of similar magnitude, as depicted in the schematic illustration combining two linear static pressure fields. In contrast, plane B_x results for TVF2 show mainly horizontal contour lines, indicating that the radial gradient is the primary contributor, leading to delayed migration, and hence weaker passage vortices.

Figure 10(b) shows the pitch angle distribution at the exit of TVF2. The vertical accumulated contour lines indicate a pronounced UPV in the splitter passage (C1). In the lower span region, the red zones (P1–P3) show a radially inwards fluid migration due to the LPV of the corresponding passages.

The most notable and unique flow feature at the exit of TVF2 is the low p_t-region spread around the circumference from the hub to ∼5% span. The circumferentially deflected flow (low axial velocity component) passed the airfoils' trailing edges near the hub faces an axial diffusion as a result of the area growth caused by the duct inclination (mean line radius increase). These lead to reduced axial momentum and ultimately over-turning of the flow in that region. The CFD has predicted this phenomenon, which was also confirmed by an oil flow visualization test, as illustrated in Fig. 12. The oil flow test was performed by injecting white paint into the L-HF cavity (Fig. 3). Moreover, this flow behavior was also found to be insensitive to HP-PFR and LP-PFR changes.

Figure 13(a) shows the normalized pressure loss at the exit of TVF2 for the 0% PFR case. The data in this figure were taken using two independent measurements in plane C—FHP measurements contribute to the data above ∼8% span while data closer to the hub is obtained with a flattened pitot probe.

Most of the loss occurs in the low-span region near the hub, in a zone that is uniformly spread around the entire circumference and extends up to about 15% span. At ∼80% span all passages contain a loss core (L4, L5, and L6) due to the UPV. The loss core L5 can be observed with an extended impact among the others. It is found that the HP-vane modulated structure from the HP-rotor impinges on the splitter’s leading edge (Fig. 5. 5R/5V). This reinforces the PS-leg of the horse-shoe vortex of the PS-splitter, which then migrates toward the SS of the SS-splitter as it passes through the passage and amplifies the UPV (distinctive vortex core; Fig. 10(b), structure C1).

Figure 13(b) shows the differential plots of the total pressure at the TVF outlet referred to as the 0% PFR case. The methodology of the plot is identical to the above introduced Fig. 9(b).

The differential plot of 50% an increase of losses at low-span positions for all passages (D4–D6) due to the LPV. However, the structure D6 located in the PS-passage is found to be the most pronounced among the others. The schematic illustration of the hub surface streamlines in Fig. 11 in combination with the static pressure of B_x indicates an unusual circumferential pressure gradient, forcing fluid migration from SS to PS (negative incidence) leading to the formation of a vortical structure V1 that lifts off where the streamlines accumulate. Before it enters the PS-passage this vortex interacts with the PS-leg of the struts HSV, which is found to be responsible for the enhanced losses.

For 100% PFR one can find the same structures D4–D6, but strengthened with respect to the 50% PFR case. Besides, other zones of increases losses appear. L5 is found to be amplified (bottom lobe) and on the other hand, extends circumferentially toward the PS-splitter (marked Y). Further, all wakes imply more loss (W1–W3), but in fact, the flow angles have slightly changed, leading to a different location of p_t-wake characteristics, thus a misinterpretation based on the applied methodology utilizing differential plots with positive range only.

Ultimately, the 200% PFR plot shows a widely similar behavior compared with the 100% PFR. The increased PFR has led to a further amplification of the LPV-related loss cores D4–D6, in particular D6. Due to the injected purge, the boundary layer at the hub is found to be thicker. This allows the formation of a strengthened vortex V1 (description under 50% PFR), which enhances the losses due to the interaction with the PS-leg of the struts HSV. In the upper part of the passage, structure L5 is found to be enlarged. Further, the expanded region observed for 100% case is enhanced (lobe Y). The increased PFR pushes the HP-rotor related structures more toward midspan. Consequently, the later interaction with the splitters LE affects wider zones of the splitter passage. The perceived rise of loss in the wakes (W1–W3), is found to be a misinterpretation caused by the methodology, as described for the 100% PFR.

The discussion of the pressure loss comparison will be based on the radial distributions of the pressure loss illustrated in Fig. 14 and the single number value normalized by the corresponding 100% PFR case shown in Fig. 15 containing the dataset of the PFR derivatives and the incidence variations.

The radial line of TVF1 indicates a strong off-design dependency up to ∼50% span, which is also the region that contributes most to the total pressure loss. Besides PFR impact, one must note the difference between the 100% PFR and the positive incidence case between 50% and 80% span. Since the lines collapse everywhere else, it implies that positive incidence enhances the circumferential pressure gradient and, ultimately, the downwash-associated UPV.

The sensitivity of TVF2 to purge and incidence derivatives is shown to be different from that of TVF1. For this TVF, most of the pressure loss occurs in the low-span region near the hub where the low-momentum flow exists (see Sec. 3.3). Moreover, TVF2 appears to be largely insensitive to PFR changes over almost the entire span, except between ∼10% and ∼30% span. In this region, the cases of 0% and 200% PFR are shown to have an impact. This region is also the one found to be affected the most in the differential plots (Fig. 13(b)).

In general, TVF2 appears to be more robust to off-design due to a larger acceleration in the duct (Fig. 2, compare AR_SW). Consequently, the inflow conditions have a less distinct impact compared to TVF1.

Figure 15 depicts the 1D loss development with increasing PFR for the two TVF designs normalized to the corresponding 100% PFR case. As the purge flowrate increases the duct loss also increases for both TVFs but in a non-linear fashion in line with the investigations of Merli et al. [24]. Following what was previously discussed, TVF2 shows less sensitivity to increases in PFR than TVF1.

Moreover, the positive incidence case of TVF2 shows an insignificant higher pressure loss than the 100% PFR (ADP), thus underlining the robustness also in terms of incidence. Conversely, the +α case of TVF1 demonstrates a 2.5% increase of loss relative to the 100% PFR, emphasizing the sensitivity to positive incidence as described in Sec. 3.2.

The last subsection focuses on the outlet flow conditions that ultimately affect the efficiency of the LPT. The outflow quality is assessed by the quantities of the Mach number normalized by its mean and the relative flow angles (yaw and pitch), referred to as the midspan metal angle and the mean duct inclination, respectively depicted in Fig. 16. Since the uniformity of those quantities is crucial for high-efficiency operating LP-rotor, the evaluation is based on three parameters: min/mean, max/mean, and standard deviation, as shown in Table 5 for 100% only since the changes due to off-design operating points are neglectable.

The distinctive difference between the two TVFs can be observed in the variation of the flow angles. The yaw angles min/mean and max/mean show no substantial variation. However, the standard deviation leads to the conclusion that the TVF2 outflow faces less non-uniformity. In general, both setups yaw angle distributions do not show a big difference between strut and splitter(s).

The opposite trend is noticeable for the pitch angle. While max/mean appears to be in the same order of magnitude, the min/mean of TVF1 is highly affected by the pronounced downwash, resulting in even negative values. Consequently, the standard deviation is one order higher compared to TVF2. Moreover, the pitch angle of TVF2 shows over almost the entire circumference a good agreement with the mean duct inclination, except the PS-passage, where it tends to pitch up.

The Mach number alternation of the setups suggests the same findings as for the flow angles but not that pronounced. The splitter wake of TVF1 appears to have approximately the same impact compared to the splitter wakes of TVF2. On the other hand, the strut wake of TVF1 shows a substantial velocity wake, which is clearly not favorable for the upcoming LP-rotor. While for TVF1, one can clearly identify strut and splitter, this is not the case for TVF2, as the wakes occur all in the same order of magnitude.

This paper explores the impact of various inflow conditions on two promising TVF design approaches, characterized by differences in splitter count and the inlet-to-outlet speed ratio, representing acceleration. The designs under comparison share the same HPT, ensuring similar inflow conditions for a fair assessment.

In the case of TVF1, the exit flow-field is characterized by a large low total pressure region adjacent to the strut suction side caused by the migration of the upper passage secondary flow vortices to low spans near the hub and their interaction with the lower passage vortices. This low-pressure region is found to be the primary contributor to the pressure loss through TVF1. The large migration and associated intensification of the upper passage vortices in this duct is due to the positive incidence at the inlet and to the low aspect ratio airfoils which promote the early formation and enhancement of secondary flows, and the duct bends which impose strong radial gradients and associated radial migrations of these features. For an engine designer an important consideration would be to control the aforementioned drivers, should the engine architecture allow such flexibility.

For TVF2, the “sensitive” region at the exit of the duct is found to be near the hub, where a circumferential band of over-turned flow exists. This region is because of a combination of a low axial momentum flow due to the high turning imposed by the duct's airfoils and by the axial diffusion resulting from an area increase at the exit of this TVF. This phenomenon was also found to be insensitive to off-design operations and LPT purge flowrates.

Regarding loss contribution, TVF1 is mainly driven by the flow feature on the strut's SS influenced by secondary flows. In contrast, for TVF2, the primary contribution can be traced back to the low-momentum band at the hub. In the case of TVF1, the identified drivers of pressure loss in the reference case were found to be amplified and extended for off-design operation.

Conversely, the impact of different PFRs in TVF2, was found to be non-identical to the identified pressure loss zones of the reference case. Changes manifest in the low-span regions and can be assigned to the lower passage vortices. Moreover, the LPV of the PS-passage was found to be responsible for the most losses in that zone. TVF2 exhibits a more uniform outflow and demonstrates broad insensitivity to off-design operation compared to TVF1, a desirable feature for the low-pressure rotor in terms of efficiency. The flow uniformity can be linked to the presence of the second splitter.

In addition to aerodynamic considerations, it is emphasized that various parameters beyond aerodynamics affect ITDs design. Parameters such as part count (splitters) play a crucial role, considering that a TVF is a hot gas component, leading to different cooling requirements, manufacturing complexities, and maintenance costs.

The project leading to this paper has received funding from the Clean Aviation Joint Undertaking under the European Commission’s Horizon Europe research and innovation program under grant agreement OFELIA No. 101102011. The authors would like to thank Y. Fuchs, A. Salah, and G. Bozzo for their contribution to operating the test facility, as well as GE Aerospace for the permission to publish this paper.

There are no conflicts of interest.

The authors attest that all data for this study are included in the paper.

r =

radius (m)

A =

area (m²)

L =

length (m)

$m˙$ =

massflow (kg/s)

α =

yaw angle (deg)

γ =

pitch angle (deg)

σ/std =

standard deviation

Θ =

theta (deg)

ADP =

aero design point

ALF =

aft looking forward

AR =

area ratio

CFD =

computational fluid dynamics

DW =

downwash

FHP =

five-hole probe

FWD =

forward

HSV =

horse-shoe-vortex

HPT =

high-pressure turbine

ITD =

intermediate turbine duct

LPT =

low-pressure turbine

LPV, UPV =

lower/upper passage vortex

PFR =

purge flowrate

PS, SS =

pressure/suction side

pt, ps =

total/static pressure (Pa)

Re =

Reynolds number

ref =

reference

TCF, TVF =

turbine center frame, turbine vane frame

TLV =

tip leakage vortex

UW =

upwash