Impact of Turbine-Strut Clocking on the Performance of a Turbine Center Frame

Efficiency is probably the most desired performance parameter in modern turbofan engines. Efforts of increasing the efficiency of an engine have traditionally concentrated on improving the efficiency of the individual components in isolation. However, this approach has only led to incremental loss reduction in today’s state-of-the art engines. To achieve significant efficiency improvements, a system approach is needed whereby the cross-influence of all of the various components in a system is taken into consideration to increase the efficiency of a single component. Turbine transitions ducts, in this case the turbine center frames (TCFs), represent the interface between the last high-pressure turbine (HPT) stage and the first low-pressure turbine (LPT) stage. The performance of these transition ducts is particularly affected by the interactions of the TCF with the upstream HPT and downstream LPT. Further TCF improvements can be achieved only based on system optimization studies including the components upstream and downstream. A review of several investigations characterizing the flow behavior of a TCF downstream of an HPT was compiled by Göttlich [1].

More recently, Zerobin et al. [2] published a study investigating an engine-relevant combined HPT–TCF–LPT vane test configuration, including HPT purge flows. Besides revealing the contributions of the HPT purge interaction mechanisms with the main air to the TCF pressure loss and confirming the findings from Jenny et al. [3] and Regina et al. [4], Zerobin et al. postulated a dependence of the TCF pressure loss on the HPT vane to TCF strut clocking position. The flow field downstream of the TCF was acquired over more than one TCF pitch, including two TCF strut wake regions. Figure 1 published in the study by Zerobin et al. [2] presents the acquired flow field downstream of the TCF. A comparison of the total pressure loss obtained by evaluating the different flow regions (strut I, core flow, strut II in Fig. 1) revealed significant differences in the two strut wake region. The total pressure loss difference between strut wake I and strut wake II was between $6%$ and $20%$ depending on the purge flowrate. The difference between these two strut regions is related to the upstream position of the HPT vane. As the number of HPT vanes was not a multiple of the number of TCF struts of this setup, the flow disturbances caused by the upstream vane structures interacting with the strut were different on each strut.

Conversely, this would mean that a similar HPT–TCF–LPT vane test configuration with an integral multiple of HPT vanes with respect to TCF struts would achieve the same total pressure loss in all TCF strut regions. By adjusting the circumferential position of the HPT vanes with respect to the TCF struts (clocking), the pressure loss could eventually be reduced to the lower level observed in strut wake region II (Fig. 1) or even further. This would result in a reduction in total pressure loss for the entire HPT–TCF system.

Clocking of turbine stages is a known approach in performance optimization of a turbomachinery system. Huber et al. [5] demonstrated the potential of clocking by showing a $0.8%$ efficiency improvement in a two-stage turbine due to clocking the second-stage vane. Eulitz and Engel [6] numerically investigated the effects of turbine clocking and concluded that clocking positions where the wake from an upstream stator is convected through the mid-passage result in higher losses.

A time-resolved numerical study by Arnone et al. [7] in a multistage LPT showed a $0.7%$ efficiency variation due to clocking. Reinmöller et al. [8] conducted hot wire and pneumatic probe measurements and showed that clocking resulted in a $1%$ relative efficiency variation at midspan. Their results were also supported by numerical simulations.

In addition to efficiency improvement, clocking can also influence the time-resolved characteristics of a turbomachinery system, resulting in a reduction of noise and vibrations. Davis et al. [9] investigated a transonic HPT with a downstream transition duct and LPT vanes and addressed the interaction among the rotor blade, the intermediate turbine duct, and the second vane. Gadea et al. [10] studied the influence of clocking on the time-resolved pressure field of a second vane tested in a 1.5-stage HP turbine without a transition duct. It was shown that the optimum clocking position for aerodynamics is not the optimum for minimum unsteady force. Schennach et al. [11] investigated a transonic HPT and found that the pressure fluctuation amplitude decrease by roughly 38% at the leading edge.

Although turbine clocking is extensively addressed in the existing literature, only a few publications concerning modern high bypass engines are available. The purpose of this study is to address this shortcoming, by investigating the effect of clocking in an engine-representative HPT–TCF–LPT vane configuration, including HPT purge flows. The impact of turbine-strut clocking on the aerodynamic performance of the TCF and the time-resolved TCF exit flow is shown, based on both, experimental and numerical data.

The transonic test turbine facility at Graz University of Technology is a continuously operating cold-flow open-circuit turbine test plant. The test rig is driven by pressurized air supplied by a 3 MW compressor station. A three-stage radial compressor is connected to the HPT and delivers additional air mixed with the air from the compressor station to increase the overall mass flow. Detailed information on the design of the HPT rig can be found in the study by Erhard and Gehrer [12]. More details on operating the rig are presented in the study by Neumayer et al. [13].

The 1.5 stage test setup itself, shown in Fig. 2, consists of an uncooled, HPT stage with an unshrouded rotor (aerodynamically representative of a second-stage HPT engine), followed by a turbine center frame, an S-shaped duct with nonturning struts, and an LPT vane row. As mentioned earlier, an integer ratio between the HPT vanes and TCF struts is needed to achieve identical TCF passage flow characteristics for all TCF strut passages. Therefore, an HPT stage with a vane-to-TCF-strut ratio of 4 was designed.

For a more engine-representative HPT exit flow, this turbine stage is equipped with purge mass flows, as illustrated in Fig. 2. Four purge flows are feeding the cavities upstream and downstream (labeled FWD and AFT, respectively) of the HPT around the full annulus. Mass flow and temperature are set individually for all four purge flows. The normalized values are listed in Table 1. More details on the development of the secondary air system are given by Steiner et al. [14].

All tests were performed at the aero design point, characterized by the operating conditions listed in Table 2. To investigate the impact of HPT–TCF clocking, the circumferential position of the HPT vanes in relation to the TCF struts was changed between the test runs. The vanes were mounted on a circumferentially traversable ring, providing a simple and fast way to adjust the clocking position. Note that the test conditions were always set with the same clocking position, and the clocking position to be investigated was set immediately before the measurement. This approach is used to minimize any variation between the tests and to increase the comparability of the results. In total, four different clocking positions were investigated. Clocking position C0 represents the nominal position. Clocking positions C1–C3 are equally distributed over one HPT vane pitch, leading to a circumferential difference between C0 and C1 of $25%$⁠, between C0 and C2 of $50%,$ and between C0 and C3 of $75%$⁠, with respect to one HPT vane pitch.

Five-hole-probe(5HP) area traverses were performed upstream and downstream of the TCF. In plane B (Fig. 2), immediately downstream of the HPT, a so-called pseudotraverse was performed due to limited space to implement a real traverse. The pseudotraverse means that the probe is fixed at a single circumferential position (mid pitch), and the HPT vanes are traversed over more than two HPT vane pitches. More details on this approach are given by Faustmann and Göttlich. [15]. The axial distance between the measurement plane B and the HPT rotor blade is normalized by the axial length of the HPT rotor blade and measures 0.7.

In plane C (Fig. 2), at the TCF exit, real area traverses were performed over one TCF strut pitch. The axial distance between the respective measurement plane and the strut trailing edge is again normalized by the axial length of the HPT rotor blade and measures 1.1.

Two different 5HPs had to be used due to the varying wall slope of the TCF. A 90 deg probe was used upstream and a prepitched probe downstream of the TCF. Both probes share the same head geometry and were manufactured and calibrated by the Institute of Jet Propulsion and Turbomachinery, RWTH Aachen University.

Table 3 presents the measurement uncertainties of the 5HP. These values contain the errors due to the polynomial approximation and the errors of the pressure and temperature transducers. The downstream probe was traversed in the circumferential and radial direction to predefined measurement points (between 800 and 1400 point per area traverse). At each location, the probes were rotated around its axis (quasi-yaw axis) to point the head into the mean-flow direction. This approach ensured that the probe was within the yaw angle calibration range at all times. 5HPs were equipped with a shrouded thermocouple, allowing a simultaneous measurement of total temperature. Pressure and temperature signals were measured for 5 s at each measurement point (excluding settling time), and the signals were acquired by PSI 9116 multichannel pressure transducers and National Instrument NI 9214 thermocouple modules.

Fast-response aerodynamic pressure probe (FRAPP) measurements were performed downstream of the TCF in the measurement plane C. An area traverse was performed using a cylindrical single-sensor probe equipped with a miniaturized piezo-resistive sensor (Kulite XCE-062). The probe was operated as a virtual three-hole probe, enabling an evaluation of unsteady Mach number, yaw angle, static, and total pressure (see Kupferschmied et al. [17]). The probe was aligned in the flow direction at each measurement location based on the yaw angle determined by the 5HP. The FRAPP was calibrated in a Mach number range of 0.2–0.8 and the transfer function was obtained in a shock tube. Details about the FRAPP design and calibration are provided by Persico et al. [18]. After digital compensation, the probe bandwidth is extended up to 80 kHz. The expected uncertainty is equal to $±0.5%$ of the kinetic head for the pressure measurements. At each measurement point, the FRAPP signal was acquired together with the trigger signal of the HPT shaft for 2 s with a sampling rate of 500 kHz.

This approach results in an amplitude spectrum over the azimuthal mode order for each radial measurement position (Eq. (10)). A quantification of the amplitudes of the individual azimuthal modes can now enable a ranking of the HPT–stator–rotor–TCF strut interactions by their impact on the unsteady flow field downstream of the TCF.

Unsteady computational fluid dynamics (CFD) simulations have been conducted by the industry partner (MTU Aero Engines AG) using TRACE (Turbomachinery Research Aerodynamic Computational Environment). The unsteady Reynolds-averaged Navier–Stokes (URANS) solver was developed at the German Aerospace Center (DLR) in collaboration with MTU Aero Engines AG. Details can be found in Franke et al. [23] and Nürnberger and Greza [24]. TRACE is based on a cell-centered finite-volume approach. The two-equation turbulence model from Wilcox and the stagnation point anomaly fix from Kato and Launder [25,26] have been used. A structured multiblock mesh including fillets has been created with y + =1 on all wetted surfaces enabling the use of low-Reynolds wall modeling. Figure 3 presents the CFD domain. The cavities of the test vehicle were meshed separately and attached to the flow path using zonal interfaces are colored in red and green in Fig. 3. The inlet domain is set upstream of the HPT vanes, at the same location used for total pressure and total temperature rake measurements in plane A (Fig. 2). These measurements (total pressure, total temperature) combined with inlet turbulence set the inlet boundary conditions for the CFD calculations. Static pressure distributions obtained from the 5HP measurements and wall static pressure taps were used to set the exit boundary conditions in plane D, downstream of the LPT vanes (Fig. 2). The mesh was generated according to the industry partner’s standards. This approach leads to a very good agreement between CFD and measurements. Particularly, the static pressure distribution along the TCF struts is in line with the measurements, and therefore, no separate grid convergence study was performed for the mesh used in the URANS simulations. The total number of cells, along with the most important mesh criteria, are summarized in the Table 4. All values refer to the unsteady CFD simulations, modeling a 60 deg flow passage. Smallest and largest angles were found in the tip fillet of the turbine center frame strut, in isolated places, and the total fraction of these cells compared to the total number of cells in the fillet is vanishingly small.

Only unsteady CFD calculations will take the HPT-clocking impact into consideration (no mixing planes). For the given investigation, TRACE makes use of the equal pitch unsteady method by simulating only a fraction of the complete test setup. To allow for the best possible agreement between the experiment and simulation, an identical postprocessing approach for the numerical data was developed. By introducing additional surfaces at measurement planes B and C (identical to 5HP locations), flow quantities have been calculated for each time-step of the unsteady CFD simulation. In the next step, the snap shot CFD data is interpolated on the measurement grid of the 5HP followed by time averaging the CFD data. The snap shots have been saved after the unsteady CFD simulations were converged (after seven rotor periods).

The following sections will present the impact of the clocking on the TCF flow. Relevant HPT structures will be highlighted and their interaction with the TCF struts will be presented inside the TCF and particularly at the TCF exit. Finally, their impact on the TCF pressure loss will be quantified.

Figure 4 shows contour plots of total pressure coefficient c_pT and streamwise vorticity ω_sw at the HPT exit (plane B). Data corresponding to experimental measurements (5HP data) are shown at top of the figure, and data corresponding to CFD simulations are shown at the bottom. In all cases, the view is aft looking forward, as shown in Fig. 2.

The flow field is dominated by the HPT secondary flows, most notable in the streamwise vorticy contours. Namely, the passage vortices of the stator or nozzle (LPV_N, UPV_N), the passage vortices of the rotor (LPV_R, UPV_R), and the tip leakage vortex are dominating the flow field. Negative values of the streamwise vorticity, colored in blue, indicate a rotation of a flow structure in the same direction as the rotor. In addtion the sense of rotation is indicated by the arrows next to the colorbar. In the absolute frame of reference, downstream of the turbine, rotor-induced structures appear as circumferential bands. Stator structures appear at fixed circumferential positions modulating the rotor structures.

The LPV_N and the counter-rotating peak value of LPV_R are forming a vortex pair convected through the duct as a characteristic streak. The previous studies (Zerobin et al. [2] or Sterzinger et al. [28]) mentioned these streaks as purge flow streaks because they are the relevant structures entraining the HPT purge air and transporting it downstream.

The 5HP data and CFD are generally in good agreement—characteristics and locations of the flow structures, as well as the absolute levels, are comparable. The CFD is predicting small variations between the two investigated HPT vane pitches, visible in the c_pT contour plots. This variation is not captured by the 5HP pseudotraverse and can be linked to the upstream influence of the TCF struts.

Figure 5 displays the numerically obtained contour of c_pT upstream and downstream of the TCF. Inside the TCF duct, the aforementioned purge flow vortices are displayed.

These streaks are located in the lower span region (up to 30% span) and can either propagate through the core region, such as the ones in Fig. 5 for the baseline clocking position C0, or interact with the TCF strut. As the LPV_N, located at a distinct circumferential location (Fig. 4), contributes to these streaks, the relative position between the HPT vane and the TCF strut (clocking position) determines the circumferential location at which these streaks propagate through the TCF and particularly whether they will interact with the TCF strut or not. Defined by the ratio between HPT vanes and TCF struts in the investigated setup, four purge streaks are present in one TCF passage.

Figure 6 presents the distributions of the static pressure coefficient c_p along the TCF struts for two clocking positions, C0 (nominal) and C2 (vanes rotated by 50% HPT vane pitch). Experimental (with symbols) and numerical results (lines without symbols) are presented for three different span heights (15%, 50%, and 85% span). As mentioned earlier, the terms rotor-wake leeward (dashed lines) and windward (solid lines) are used to label the two strut surfaces.

As the TCF duct diffuses the flow, static pressure is increasing along the TCF chord. This is more pronounced at higher spans. At 85% span (top diagram in Fig. 6), the pressure distributions for both clocking positions are on top of each other. This indicates no influence of clocking in this region. At mid span (middle diagram in Fig. 6) and particularly at 15% span (bottom diagram in Fig. 6), the changes induced by clocking are more evident. At design conditions, the TCF strut was locally at increased incidence at 15% span, as indicated by the pressure difference between RWL and RWW side in the first part of the TCF chord. Changing from baseline clocking position C0 to clocking position C2 leads to a higher incidence angle on the TCF strut at 50% span, whereas the incidence angle at 15% span is reduced. The CFD predicted no interaction between the purge streaks and the TCF strut for the clocking position C0 (Fig. 5). Therefore, the differences observed in the static pressure distribution along the TCF strut surface can be related to an interaction of the streak and the TCF strut for the clocking position C2.

In general, the CFD is in very good agreement with the measurements and can be used with confidence to get additional information toward the leading and trailing edge of the strut. The largest deviations can be observed at $15%$ span on the RWW side of the TCF strut. At higher span (⁠$50%$ and $85%$⁠), the agreement between CFD prediction and measurement is better.

More details can be observed from the experimentally and numerically data downstream of the TCF, presented in the following section.

The flow field downstream of the TCF (plane C) is experimentally captured by conventional 5HP area traverses. As the HPT vane number is a multiple of the number of TCF struts, the periodicity is one TCF pitch. To confirm this, the flow field was initially measured over 1.2 TCF strut pitches including two TCF strut wakes. Both TCF strut regions showed the same qualitative behavior. To decrease measurement time, only one TCF pitch was acquired for subsequent test runs.

Figure 7 presents the stream wise vorticity ω_sw at the TCF exit for two clocking position C0 and C2. Data corresponding to experimental measurements (5HP data) are shown at the top of the figure, and data corresponding to CFD simulations are shown at the bottom.

The traces of the purge flow vortices are still present and highlighted with the labels B1–B4 in Fig. 7. For the nominal clocking position, all four purge flow streaks are present. In the structures B3 and B4, still both counter-rotating vortices can be observed (LPV_N and LPV_R). The structures B1 and B2 are weakened downstream of the TCF. The vortex related to the LPV_N (rotating against the direction of the HPT rotor, colored in red) is almost mixed out, whereas the counter-rotating vortex, related to peak value of the LPV_R (colored in blue, rotating clockwise), is enhanced and spread out.

The loss generation caused by these vortices is shown in Fig. 8. Total pressure coefficient c_pT at the TCF exit for two clocking positions C0 and C2 is shown. For the baseline clocking position low c_pT regions (B1–B4) are visible at 0.05, 0.3, 0.55 and 0.8 Θ/Θ_TCFstrut, respectively, in Fig. 8. The circumferential locations are corresponding to the locations were the purge flow vortices were observed in the streamwise vorticity contour (Fig. 7).

For the clocking position C2 in Figs. 7 and 8, the vanes were moved by 50% of one HPT vane pitch with respect to the baseline clocking position. With the known ratio, HPT blades to TCF struts of four, the theoretical circumferential shift of structures related to the HPT vanes downstream of the TCF is 0.125 Θ/Θ_TCFstrut in clockwise direction. This is only a simple estimation, not including the TCF strut interaction and the pressure gradients present in the duct. Nevertheless, this estimation is acceptable since the structures B1, B3, and B4 can be observed at approximately 0.18, 0.68, and 0.93 Θ/Θ_TCFstrut, respectively, in Figs. 7 and 8. Particularly interesting is the absence of structure B2 in the flow field downstream of the TCF for clocking position C2. This structure would be expected at approximately 0.43 Θ/Θ_TCFstrut, very close to the TCF strut located at 0.5 Θ/Θ_TCFstrut. Therefore, it can be assumed that the streak, causing the low total pressure region B2, undergoes a strong interaction with the TCF strut. The vortex pair integrates in the strut flow and can no longer be observed in the streamwise vorticity (Fig. 7, C2). However, the flow field in the TCF strut region is changed by this interaction.

For the clocking position C2, the high total pressure region (marked with C) on the right-hand side of the TCF strut is present for both clocking positions and slightly enlarged compared to the baseline clocking position C0. This indicates a freestream region on this side of the strut and no purge streak interaction.

On the left-hand side of the TCF strut, in Fig. 8, the low total pressure region (labeled with A) is pushed away from the strut, probably by a radial fluid migration caused by the interaction of the purge streak and the TCF strut surface and implies that the purge streak B2 impinges on the left side of the TCF strut (rotor-wake leeward side), likely caused by some HPT exit swirl pushing the fluid toward the rotor-wake leeward side of the strut. The low momentum fluid caused by the streak B2 is transported to the TCF outer surface due to the pressure gradient in the TCF duct (Göttlich [1]) and mixed with the already present low total pressure zone A (Fig. 8) at the outer endwall of the TCF. Therefore, it can be assumed that for C2, the number of loss cores is reduced. Instead of disturbing a high-pressure region, as for the baseline clocking position C0, one purge streak is pushed into a low momentum region, which is always present independent from the clocking position. This could be an indication of the advantage of clocking position C2 over C0. A mixing between a high momentum fluid and a low momentum fluid causes larger losses in comparison to the mixing of two low momentum zones.

Figure 9 presents the Mach number distribution at the TCF exit. In general, the Mach number distributions confirm the mechanism explained earlier. It can be observed that the low c_pT regions (B1–B4) observed in Fig. 8 correspond to low Mach number regions, highlighted in the same way in Fig. 9. For the clocking position C2, the structure B2 is interacting with the TCF strut. Therefore, the Mach Number is significantly higher between 0.3 and 0.4 Θ/Θ_TCFstrut. To further investigate the impact of this interaction mechanism on the downstream components, mass-averaged radial profiles will be presented next.

The URANS calculation captures the same structures and again, a good spatial agreement of all present structures is seen. All significant changes due to HPT vane clocking were captured by the CFD, and the level of total pressure coefficient (Fig. 8) and Mach number (Fig. 9) are in good agreement. Compared to the 5HP results, the absolute values of the streamwise vorticity (Fig. 7) are higher, indicating stronger vortical structures in the CFD predictions, compared to the measured results. This discrepancy could be caused by an underprediction of the mixing in the diffusing TCF duct by the CFD. Particularly in the TCF strut region at 0.5 Θ/Θ_TCFstrut, the 5HP results show lower values of ω_sw, indicating that these vortical structures from the TCF strut wake may already be mixed out, although the CFD still captures them downstream of the TCF in plane C. A consequence of underpredicting the mixing can be observed in the Mach number distributions in Fig. 9. The 5HP measurements capture a distribution with low gradients in span-wise and pitch-wise directions. The URANS calculations underpredict the mixing in the TCF, resulting in smaller structures with higher or lower values, compared to the 5HP values and stronger gradients in span-wise and pitch-wise directions.

Figures 10–12 present the circumferentially mass-averaged radial profiles, calculated using Eqs. (1) and (2). As observed in the contour plots shown earlier, the TCF strut region is of particular interest. Therefore, the TCF exit flow field is divided into different circumferential flow sectors, as shown in Fig. 7. The strut region is indicated by the blue box labeled strut region, and the core flow region on both sides of the strut is indicated by the red box labeled core flow. The borders between the two regions are not related to the geometry. The sector from 0.4 to 0.6 Θ/Θ_TCFstuts was defined as the strut region and the remaining sector as core region. This approach leads to three different profiles for each flow quantity: one for the entire TCF passage, one for the core region, and one for the TCF strut region, labeled with (a), (b) and (c) in Figs. 10 and 11, respectively. These figures present the radial profiles for two flow quantities, total pressure coefficient c_pT and streamwise vorticity ω_sw. The experimental (5HP) and numerical data are presented for the same clocking positions C0 and C2 as discussed earlier.

For the core region, the experimentally observed differences between the two clocking positions (Fig. 10(b)) are quite small although the prior discussion of Fig. 8 revealed that the magnitude of single low total pressure regions (B1–B4) changed with the clocking position. This indicates that in the core flow the total pressure loss is just redistributed, and no benefit of the overall total pressure loss can be achieved by clocking the HPT vanes in the core region. The numerical results show higher difference of the c_pT profiles and a small benefit for the clocking position C2, but again no major changes can be observed in this region.

Significant changes caused by the HPT-clocking positing are expected in the TCF strut region, and therefore, Fig. 10(c) presents the radial profiles for this region. The 5HP profile for the clocking position C2 shows higher c_pT values between 10% and 45% span. Above 50% span, the c_pT profile for the C2 clocking position is below the baseline profile (C0). Nevertheless, this diagram does not show specifically whether an overall pressure loss was achieved by changing the clocking position, but a more uniform exit profile can be observed for the clocking position C2. This could be beneficial for downstream LPT stage.

The CFD shows the same trend at lower span (below 50%), but larger discrepancies in the tip region.

The radial profile of the entire passage (Fig. 10(a)) is basically the sum of the two previously discussed profiles. Note that only approximately 20% of the entire TCF passage mass flow passes through the strut region. In Fig. 11, the radial profile of the streamwise vorticity is presented. The largest differences related to the clocking can again be observed in the strut region (Fig. 11(c)), whereas the streamwise vorticity in the core region (Fig. 11(b)) is less affected by the clocking. The CFD prediction (Fig. 11) is generally following the same trend. Highest deviation between experimental and numerical data can be observed at 60% span. This is could be caused by different magnitudes of streamwise vorticity observed in the TCF strut wake (Fig. 7).

Figure 12 presents the radial distributions of the Mach number. Again, it can be observed that largest differences induced by the HPT clocking position are in the TCF strut region (Fig. 12(c)). This observation supports the aforementioned findings, i.e., that the effects of HPT vane clocking in the flow field downstream of the TCF is constrained to the TCF strut region. The numerical simulations capture the same overall trend and particularly the degradation of the Mach number from hub to tip. In general, the absolute level of the Mach number is slightly higher in the calculation, and differences between the clocking positions C0 and C2 are more evident in the CFD results, predicting a higher TCF exit Mach number for the clocking position C0.

Up to now, only an assumption about the behavior of the TCF pressure loss was presented. To support this assumption and quantify the benefit, this section discusses mass-averaged single values of the total pressure loss for the different clocking positions. Rake-based data are also leveraged for this discussion. As the measurement time for rake measurements is significant shorter, data were acquired for all four clocking positions (C0, C1, C2, and C3). Figure 13 presents the pressure loss observed in three different ways for all clocking positions. DP/P is normalized by the result for the baseline clocking positions C0.

Independent from the selected measurement technique, the lowest total pressure loss is observed for the clocking position C2 and the highest pressure loss is observed for C0. The values of total pressure loss for the clocking positions C1 and C3 are similar and between the ones for C2 and C0. This result strengthens previous explanations, stating a reduction of the number of loss cores reduces the pressure loss. The more the purge streaks are pushed into already existing low-pressure regions to lower the overall pressure loss. The reduction of total pressure loss calculated with the 5HP is in good agreement with CFD results and a relative reduction of approximately $5%$ can be achieved by clocking position C2 compared to the baseline position C0.

The pressure loss calculated from the rake data show the same qualitative trend, but the achievable benefit is smaller. The coarser resolution of the rake traverse, compared to the 5HP traverse and the different averaging procedure of the rake data (area average, Eq. (6)), could cause this difference.

Figure 13 (second axis) additionally presents the normalized turbine efficiency η_Turb calculated according to Eq. (11) using the numerical data. No dependency between the turbine efficiency itself and the HPT vane- TCF strut clocking position can be observed, leading to the conclusion that the observed TCF pressure loss reduction can be beneficial for the entire HPT–TCF system.

Figures 14 and 15 present the dependency of the amplitude of the individual azimuthal modes of the normalized total and static pressure of the clocking position, measured downstream of the TCF in plane C. Therefore, the data of two clocking positions, C0 (black bars) and C2 (orange bars) and three different span heights, close to the hub at $12.5%$ span, at midspan $50%$ span, and close to the endwall at $97.5%$ span is presented. Note that, for the setup investigated in this article, the number of the HPT vanes is a multiple of the TCF struts (ratio four to one). Therefore, different interactions can be present in the same mode. Downstream of the TCF in plane C the rotor–stator interaction m_B−V and the rotor–TCF interaction m_B−4*S will be expressed by the same mode number. For simplicity, only the name of the simplest interaction will be used (e.g., m_B−V for the aforementioned case) although this mode could additionally contain different interactions. For confidential reasons, absolute mode numbers cannot be shown. Modes of interest are therefore labeled in Figs. 14 and 15. Modes spinning in the same direction as the rotor are represented by positive-mode numbers, whereas modes spinning in opposite direction with respect to the rotor are represented by negative mode numbers.

In general, a dependence between the clocking position and the unsteady flow field can be observed at all span height in both the total and the static pressure. Depending on the clocking position, the amplitude of individual modes is changed significantly. The HPT rotor–stator interaction mode m_B−V is still present downstream of the TCF in total and static pressure. Its amplitude is varying along the span and more pronounced at lower span. The HPT vane clocking is altering its amplitude. The changes in the total pressure (Fig. 14) are most pronounced at higher channel heights (⁠$97.5%$ span), whereas at midspan (⁠$50%$ span), the amplitude of mode m_B−V is unchanged. The changes of mode m_B−V caused by the clocking are larger for the static pressure. At midspan (⁠$50%$ span), the amplitude has changed approximately by a factor of two. For both investigated flow quantities, clocking position C0 shows lower amplitudes of the mode m_B−V at $50%$ and $97.5%$ span. But at $12.5%$ span, the trend is the opposite. C0 shows higher amplitudes of the mode m_B−V compared to the clocking position C2. The purge streaks are caused by the HPT stator interaction and can therefore somehow be linked to this mode. The radial redistribution of these purge streaks observed in the 5HP and CFD Data (Fig. 8) could therefore explain the opposed trend observed in the modal decomposition of the unsteady total and the static pressure distribution. The rotor–stator–TCF strut interaction m_B−V−S draws a similar picture, but for this mode, the changes in the amplitude are comparable for the total and the static pressure. In Fig. 14, the changes of the total pressure amplitude are more pronounced in the endwall region (⁠$12.5%$ and $97.5%$ span) compared to midspan (⁠$50%$ span). The amplitude of the static pressure mode in Fig. 15 is altered by the clocking mainly at midspan (⁠$50%$ span). The amplitude is lower for the clocking position C0. At lower channel height (⁠$12.5%$ span), the opposite trend can be observed, C0 showing higher amplitudes. The link to the mechanism causing these variation of the amplitudes is not as obvious as for the aforementioned mode, but it is assumed that the different interaction of the purge streak and the TCF struts (induced by the clocking) again causes these variations.

Even the rotor mode m_B itself is altered by the clocking position. Although smaller variation can be observed for the total pressure at all channel heights (Fig. 14), the variation is larger for the static pressure. Particularly at $97.5%$ span, a lager variation can be observed in the static pressure (Fig. 15).

For all other modes shown in Figs. 14 and 15, similar observations can be found. The HPT vane clocking is definitely causing variations of the unsteady flow field of the total and the static pressure, but an overall beneficial clocking position cannot be observed.

Particularly, the static pressure fluctuations are influencing the structural dynamics of the downstream components, and in this case, at the first LPT stage, a reduction of one distinct mode accomplished by the HPT vane-TCF strut clocking can help to resolve a potential dynamic problem of the LPT.

This study presents an experimental and numerical dataset from an engine-representative HPT–TCF–LPT vane test setup with HPT purge flows.

New insights on the interaction of the HPT flow structures with the downstream TCF and particularly the impact of the turbine-strut clocking are given. The following conclusions are drawn from the discussions presented in this article.

Impact on the Steady Flow Field and the Pressure Loss: An increased efficiency of the TCF was achieved and quantified by turbine-strut clocking. A relative reduction of TCF pressure loss of approximately $5%$ was achieved by clocking of the HPT vanes. The underlying mechanism for this efficiency increase is related to the purge streaks caused by the HPT stage. If one of these streaks interacts with the TCF strut, the low momentum fluid, generated by the streak, is mixed and pushed into an already present low momentum zone. In this way, the number of loss cores in the freestream is reduced, resulting in total pressure loss benefits. In addition, a more uniform TCF exit profile in the TCF strut region, beneficial for the downstream components, was archived by the HPT vane clocking.

Impact on the Unsteady Flow Field Downstream of the TCF: The unsteady data postprocessed and presented as azimuthal modes ranked by their amplitude revealed that the turbine-strut clocking influences the unsteady behavior of the flow field downstream of the TCF. Depending on the clocking position, the amplitude of individual azimuthal modes was changed in the mode spectra of total and static pressure. Although no overall beneficial clocking position can be given, a well-defined clocking could minimize the amplitude of distinct modes and therefore help to reduce an existing dynamical problem of the HPT–TCF–LPT system.

Suggestions for Future Designs: The obvious suggestion is to change the HPT vane count and align in the described way, to achieve a performance increase described in this work. Another possible approach for achieving even more overall benefit in an engine is to counteract the performance degradation of an engine over lifetime. Existing work (Zerobin et al. [2]) described a dependence of HPT purge flowrate and circumferential location of the purge streaks (purge clocking effect). The additional fluid ejected through the aft hub cavity alters the swirl angle in the hub region of the TCF inlet and shifts the purge streaks in circumferential direction. The purge flowrate will increase during the lifetime of an engine as the seals wear out. By designing a baseline clocking position away from the optimum, the purge clocking effect can alter the circumferential position of the streaks and let them impinge on the strut surface after several hours in service, reduce the TCF pressure loss, and counteract some of the common lifetime degradation of the engine. Additional benefit of the purge streak impingement on the TCF strut surface could be related to the heat transfer on the TCF struts. As the purge streaks transport colder purge fluid, a purge streak strut interaction could help to cool the struts, to extend their lifetime or to change to weight saving materials. However, additional heat transfer measurements are needed to provide more insight into this topic.

This work has been carried out in collaboration with GE Aviation Munich and MTU Aero Engines, as part of the research project LuFo V-3 OptiTCF (contract no. FKZ 20T1705B) funded by the German ministry of industry BMWi and has additionally received funding from the Clean Sky 2 Joint Undertaking under the European Union’s Horizon 2020 research and innovation program (Grant No. 785313). The authors would also like to thank H. P. Pirker for operating the test facility.

There are no conflicts of interest.

c =

absolute velocity

m =

Azimuthal mode order

p =

static pressure

r =

radial coordinate

t =

time

A =

area

B =

number of HPT blades

R =

gas constant air

S =

number of TCF struts

T =

static temperature

V =

number of HPT vanes

$m˙$ =

mass flowrate

c_p =

static pressure coefficient

c_P =

specific heat capacity air

c_pT =

total pressure coefficient

N_c =

number of circumferential points

p_T =

total pressure

T_t =

total temperature

Ma =

Mach number

A(m) =

Fourier coefficient

DP/P =

total pressure loss

α =

yaw angle

γ =

pitch angle

η_Turb =

usentropic total-to-total efficiency

Θ =

circumferential coordinate

ω =

vorticity

Ω =

rotational speed

i =

radial index

is =

isentropic

j =

circumferential index

r =

radial

ref =

reference

SW =

streamwise

t =

tangential

x =

axial