The paper discusses the time-averaged flow of a new-concept turbine transition duct placed in a two-stage counter-rotating test turbine. As a possible architecture for the turbine transition duct of future engines, the structural vanes carrying the bearing loadings can be integrated with the first low-pressure vane row in one aerodynamically optimized wide-chord vane called the turning midturbine frame. To increase the flow uniformity and to decrease the unsteady content of the flow, a baseline turning midturbine frame is redesigned by embedding two splitter vanes into the strut passage. The discussion on the flowfield is based on numerical results obtained by computational fluid dynamics and validated by aerodynamic measurements. In particular, the splitters are seen playing a major role in suppressing the big structures generated by the struts and the secondary flows of the high-pressure turbine. On the other hand, new losses are introduced by the splitters. Such structures play a decisive role in the overall component performance; therefore, their effect should be properly understood. This work provides a deep insight into the flow physics of an embedded turning midturbine frame for next-generation aeroengines, which is seen as a promising architecture in order to compact the engine size while keeping component performance high. The present work can be considered as a first attempt to implement splitter blades into an already existing turning midturbine frame design, and its value lies in a thorough experimental and computational study on the losses and benefits of such a design.

Nomenclature
$C$	axial chord length
$C_{P}$	static pressure coefficient
$C p_{t}$	total pressure coefficient
$H$	relative channel height
$Ma$	Mach number
$P$	relative pitch
$p$	static pressure
$p_{t}$	total pressure
$T$	static temperature
$T_{t}$	total temperature
$v$	velocity
$x$	axial coordinate
$α$	yaw angle
$Δ p$	pressure drop
$ζ$	total pressure loss
$ω$	vorticity

I. Introduction

The major goal of higher efficiencies for future jet engines can be partly reached with a high bypass ratio, which leads to a larger fan. If the rotational speed is not changed, the tip velocity of the fan rotor increases, which results in higher noise emissions as well as structural problems. Therefore, increasing the fan front section demands a decrease of the low-pressure (LP) stages shaft speed or an uncoupling of the shaft speed of the LP stage and the fan. The latter can be achieved by a geared turbofan configuration. In this work, the geared turbofan is not investigated; however, a two-spool configuration with a low-speed LP turbine and a high-pressure rotor with a high rotational speed is investigated in order to reach a higher enthalpy drop per stage.

The resulting larger difference in the shaft rotational speeds leads to an increasing difference in the components’ diameters (compressors and turbines). Therefore, focusing on the turbine, the resulting S-shaped design of the diffusing transition duct between the high-pressure (HP) and LP turbines is quite important for the optimization of the overall engine aerodynamic performance and weight.

Fundamental work has been done by Dominy et al. [1] about the flows in such swan-neck diffusers, and they revealed the effects of two bends on the generation of secondary flows. Many different studies have been performed on midturbine frames with nonlifting strut blades; e.g., Norris et al. [2], Wallin et al. [3], and Arroyo Osso et al. [4] performed experimental and/or numerical investigations. Marn [5] analyzed the influences of the duct dimensions as well as the aerodynamic performance of such a setup. Recently, Göttlich [6] published a review on the present state of the research on these components.

As a further measure to reach even more weight and length reduction, the low-pressure vane row can be merged with the struts, leading to an integrated concept called the turning midturbine frame (TMTF). Such a concept is characterized by wide-chord vanes of low aspect ratios, which induce a highly three-dimensional flow at the LP rotor inlet that can also affect the vibrational response of the turbine [7]. Intensive experimental work on the flow in a TMTF was performed at the Institute for Thermal Turbomachinery and Machine Dynamics (ITTM) in a two-spool two-stage counter-rotating transonic test turbine facility. In the frame of the European Union DREAM (valiDation of Radical Engine Architecture systeMs) project, the flow in a TMTF with axisymmetric and contoured endwalls was investigated [8,9]. It was shown that the LP rotor inlet flow was characterized by large structures such as wakes and secondary flows generated by the struts. These structures contributed to a performance reduction of the following LP rotor and represented a potential source of vibration for the rotor blades. Although the aforementioned geared turbofan does not necessarily need the S-shaped duct, the idea of the TMTF can also be used for such a design and is not limited to the two-spool setup discussed in this work.

To reduce the negative effect of the few turning struts, the application of an embedded design concept was suggested. The basic idea was to merge the struts and the LP vanes in one multisplitter component. Such a concept was investigated at the von Kármán Institute (e.g., Lavagnoli et al. [10]) and at the Institute for Thermal Turbomachinery and Machine Dynamics. Spataro et al. [11,12] proposed a setup embedding two splitters into the strut channel of an existing turning midturbine frame placed between two counter-rotating turbines. Although the splitter design had the drawbacks of a higher engine weight and increased complexity, which resulted in higher costs, the advantages of the setup in regard to efficiency may offset these disadvantages. Spataro et al. already showed in steady [11] and unsteady [12] measurements the reduction of the nonuniformity of pressure fluctuations (reduced strut wakes and secondary vortices) of the embedded design. In addition to the measurements, steady [13] and unsteady [14] computational fluid dynamics (CFD) simulations have been performed at the Institute for Thermal Turbomachinery and Machine Dynamics to analyze the flow structures within the TMTF with the embedded design.

The present work gives a deep insight into the flow features of a turning midturbine frame, comparing a design with and without the splitters, and allowing an evaluation of losses within the duct. The improvement and homogenization of the flow due to the embedded design is discussed, and it is compared to the baseline case without splitters. In addition to that, new flow features are presented and a deep insight into the secondary flow phenomena close to the strut and splitter walls is given as well as a comparison of the efficiency of the LP rotor. As an outcome of this work, it is presented that this forward-looking design has an impact on losses and power output of the low-pressure turbine, and that is a promising architecture for future aeroengines. The results presented give a unique insight into the flow features of an embedded design and allow other research groups and designers to use the findings of this paper for their future TMTF design.

II. Experimental and Numerical Setup

A. Test Facility and Splitter Design

The transonic test turbine facility is a continuously operating two-stage cold-flow open-circuit plant, which consists of a transonic HP stage and a counter-rotating LP stage (a schematic drawing is shown in Fig. 1). This unique configuration allows the testing of rig inserts under engine-representative conditions. Detailed information on the design of the two-spool test rig was given by Hubinka et al. [15,16]. The main parameters of the HP stage and the LP stage ( $TMTF + LP$ rotor) as well as the operating conditions can be found in [11,13].

In Fig. 1, the meridional section of the flow channel with the turning strut and the splitter blades (in red) is shown, whereas Fig. 2 shows the three-dimensional design of the strut and splitter blades. The letters C, E, and F mark the axis-normal measurement planes in front of and behind the TMTF and behind the LP rotor. The blue lines indicate the planes used in the discussion of the CFD results.

The goal of this project is the improvement of the flow through an existing TMTF duct by the addition of splitter blades. The baseline duct without splitters has been investigated in previous works [8], showing a large vane passage vortex due to a strong negative incidence at the strut blades and extended flow separations at the trailing-edge region. To reduce these loss structures and to homogenize the flow downstream of the TMTF, the splitter blades were designed based on a numerical investigation undertaken on this specific baseline configuration of the TMTF [11] at the aerodynamic design point. The axial extension of the splitters as well as the splitter count represented a compromise between aerodynamic effectiveness and solid blockage, whereas the meridional flow channel was kept the same as for the baseline case. The axes of the two splitters were located at the relative pitchwise positions of $P = 0.33$ and $0.67$ , respectively, in order to obtain equispaced wakes at the LP rotor inlet. For each splitter seven control sections located at different spanwise locations (relative channel heights of $H = 0.00$ , $H = 0.05$ , $H = 0.25$ , $H = 0.50$ , $H = 0.75$ , $H = 0.95$ , and $H = 1.00$ ) were chosen. For each section, the same spanwise metallic angle and thickness distribution as the strut trailing edge were imposed. At the leading edge, the splitter inlet blade angle was adjusted, targeting to a zero-incidence condition. The generation and evaluation of several designs defined the final shape, which was manufactured and tested. More details of the design were published by Spataro et al. [11].

It should be mentioned that the blockage caused by the splitters could not be compensated by area ruling in the present design because the baseline TMTF channel had to be used for cost reasons so that the flow between the struts was additionally influenced by the blockage effect of the splitters. On the other hand, keeping the strut design allowed a better comparability with the original TMTF design.

B. Numerical Setup

1. Mesh Setup

A steady-state numerical investigation of the flow in the LP stage (TMTF and LP rotor) was performed. In Fig. 1, the numerical setup and the computational domains can be seen. Between the stationary midturbine frame and the low-pressure rotor, a mixing plane was placed as an interface. The inlet boundary was located at measurement plane C, and the values for the boundary condition were taken from circumferentially averaged measurement data acquired by Spataro et al. [11] (Fig. 3). The outlet boundary for the numerical investigation was placed at an axial distance of $(x_{out} - x_{{LP}_{rotor}, TE}) / C_{{LP}_{rotor}} = 6$ downstream of the LP rotor trailing edge. At this position, the static pressure was known from pressure taps mounted in the casing [9].

Fig. 3
Radial distribution at plane C of the total pressure coefficient $C p_{t}$ , yaw angle $α$ , and pitch angle $γ$ .

The mesh used for the simulation consists of about 3.2 million nodes. The mesh is a multiblock structured grid generated by the in-house mesh generator AIGrid. An O-type grid is wrapped around all blades, H-type blocks are used in the main flow regions. The refinement of the nodes toward the wall has been solved with a tangens hyperbolicus distribution. This results in a mean volume ratio (volume of the first cell off the wall divided by the volume of the cell at the wall) of the entire mesh of 1.64. The $y^{+}$ value is smaller than 1 next to the blades and the outer and inner casing. The smallest physical volume of the cells is $3.072 \cdot 10^{- 17} m^{3}$ , and the mean volume is $7.63 \cdot 10^{- 10} m^{3}$ .

To validate the mesh independence, different mesh sizes were analyzed. The number of nodes is given as a curve in Fig. 4. As a characteristic value for the mesh independence study, the mean total pressure in plane E and its relative deviation to the measured value have been analyzed. The values of the mean total pressure were normalized to the value of the “ $index factor = 4$ ” result. Figure 4 shows that meshes 5 and 6 give identical results. The results of the mesh index factor equal to four deviate slightly from the mesh index factors of five and six, so that it was chosen as a good compromise between accuracy and computational effort.

2. Numerical Solver

For the numerical simulation, the commercial code ANSYS CFX® v14.0 was used. The code uses a pressure correction scheme. The high-resolution scheme was selected for the advection as well as the turbulent numerics. The set of equations solved by the CFX solver, which describes the conservation of mass, momentum, and energy, is commonly known as the Navier–Stokes equations, and they are solved with the finite-volume method. The equations are discretized with first-order accuracy in areas where the gradients change sharply to prevent overshoots and undershoots and to maintain robustness, and they are discretized with second-order accuracy in flow regions with low variable gradients to enhance accuracy [17].

In this work, a two-equation turbulence model (namely, the Menter $k - ω$ shear-stress transport [18] model) has been used. This model combines the advantages of the $k - ε$ model in the freestream, avoiding the sensitivity of the $k - ω$ model to inlet freestream turbulence, with the advantages of the $k - ω$ model close to the wall, which can be used down to the wall without any damping functions.

The fluid air used in the numerical study is modeled as ideal gas with temperature-independent properties taken from the reference state at $p = 1013.25 mbar$ and $T = 298, 15 K$ and as listed in Table 1. The fluid is Newtonian, and the Stokes assumption is taken into account. For all other parameters, the default settings were used.

One-thousand iterations were performed, where convergence was successfully achieved after 500 iterations (see Fig. 5). The magnitude of the rms value of the residuals of the equations of continuity (P-Mass) and momentum in the three directions (U-Mom, V-Mom, W-Mom) was reduced by at least six orders.

III. Results and Discussion

In this section, the results of the numerical simulation are presented and the flow features are discussed. First, the boundary conditions influenced by the flow structures coming from the HP stage are discussed. Then, the flow evolution through the TMTF up to plane E upstream of the LP rotor inlet is analyzed.

A. Flow at the TMTF Inlet

Figure 3 shows the circumferentially averaged nondimensional total pressure coefficient $C p_{t}$ , the yaw angle, and the pitch angle from the measurements in plane C (see Fig. 1), which are prescribed as inlet boundary conditions.

The total pressure coefficient is defined as follows:

C p_{t} = \frac{p_{t} - \bar{p_{t, C}}}{\bar{p_{t, C}} - \bar{p_{C}}}

(1)

where

p_{t}

is the local total pressure, and

\bar{p_{C}}

and

\bar{p_{t, C}}

represent the averaged static and total pressure at plane C, respectively.

The main flow features that dominate the time-resolved flowfield downstream of the high-pressure shroudless turbine are secondary flows. Five-hole probe measurements performed downstream of this stage were presented by Spataro et al. [11]. Bader et al. [13] discussed the appearance of the rotor tip leakage vortex (TLV) and the rotor lower passage vortex (LPV) and upper passage vortex (UPV) (see Fig. 3). The effect of the TLV at $r / H = 0.8$ leads to less swirl and a higher total pressure close to the shroud. The LPV influences the flow in the lower half of the flow channel.

The verification of the simulation has been done with a comparison with the measurement results, as was shown by Bader et al. [13].

B. Flow Evolution in the TMTF

1. Channel Inlet to the Splitters

For a better understanding of the flow evolution, it is important to understand the main sources of the pressure gradients acting on the fluid, which can be listed as follows (see Spataro et al. [9]):

The first bend of the meridional path (upstream of the strut) generates a pressure distribution that pushes the flow toward the outer casing.
The circumferential deflection imposed to a flow in an annular configuration generates a suction-to-pressure-side gradient and a radial pressure gradient. The latter pushes the fluid toward the hub endwall, and its strength will increase with the increased turning induced by the blades (swirl effect).
The second bend of the meridional flowpath located downstream of the TMTF path induces a pressure gradient that pushes the flow toward the hub.
The three-dimensional design of the strut and splitters also influences the spanwise pressure gradients.

To follow the flow through the TMTF, several planes along the flowpath are studied in more detail: $x_{mid} / C = - 0.01$ , $x_{mid} / C = 0.26$ , $x_{mid} / C = 0.73$ , $x_{mid} / C = 0.885$ , $x_{mid} / C = 0.905$ , $x_{mid} / C = 0.98$ , and $x_{mid} / C = 1.02$ (see Fig. 1). The planes are inclined in order to be approximately normal to the main flow direction or parallel to the splitter leading or trailing edge, respectively. The coordinates of the planes refer to their midspan position.

Figure 6 shows the streamwise vorticity $ω_{SW}$ [Eq. (2)], the static pressure coefficient $C_{P}$ [Eq. (3)], and the total pressure loss $ζ$ [Eq. (4)] in these planes. The loss $ζ$ is normalized with the averaged loss at the TMTF exit plane $\bar{ζ_{E}}$ (plane E in Fig. 1). The planes are viewed against the flow direction:

ω_{SW} = \frac{v_{x} \cdot ω_{x} + v_{y} \cdot ω_{y} + v_{z} \cdot ω_{z}}{v}

(2)

C_{P} = \frac{p - \bar{p_{C}}}{\bar{p_{t, C}} - \bar{p_{C}}}

(3)

ζ = \frac{\bar{p_{t, C}} - p_{t}}{\bar{p_{t, C}} - \bar{p_{E}}}

(4)

In the following discussion, special emphasis has been laid on the evolution of secondary flow structures; Fig. 6 has to be regarded together with Figs. 7a, 7b, and 8, showing the streamlines at the hub and shroud as well as at the blade walls, to better understand the complex secondary flow. The flow in the first plane of Fig. 6 at $x_{mid} / C = - 0.01$ (close before the strut leading edge at midspan) is dominated by the structures coming from the upstream stage. They are observable in the vorticity plot as clockwise (blue) and counterclockwise (green–yellow) rotating vortices: the lower-passage vortex (structure A), the weak upper passage vortex (structure B), and the tip leakage vortex (structure C). In the middle of the flow channel, the pressure gradient from hub to shroud caused by the bend is clearly visible. This gradient and the potential effect of the strut have already caused the circumferentially evenly distributed flow from the inlet to roll up. On the other hand, the losses are still relatively uniform in circumferential direction. The effect of the sweep design can be seen in the pressure gradients just in front of the struts, pushing the flow from midspan toward the hub and shroud. As a result, four small vortices can be found in the vorticity plot (structures $E 1 / E 2$ and $F 1 / F 2$ ). These structures are the leading-edge vortices of the pressure (denoted by 1) and suction (denoted by 2) side at the upper (denoted by E) and the lower (denoted by F) part of the channel, respectively. They are caused by the interaction of the boundary-layer vorticity with the radial flow, as explained by Spataro et al. [9]. Structures $E 1 / E 2$ and F1 are also visible by a strong radial shift of the wall streamlines close to the leading edge in Fig. 7a.

Fig. 7
CFD visualization of the streamlines at the shroud.

Fig. 8
CFD visualization of the streamlines for baseline and embedded design.

The pressure distribution at the next plane $x_{mid} / C = 0.26$ in Fig. 6 shows two remarkable gradients. In the hub region, the pressure at the suction side (SS) is higher than at the pressure side (PS), which is caused by a negative incidence so that a lower passage vortex cannot evolve at this position. On the other hand, the relatively high pressure at the hub and the spanwise change in the incidence lead to a strong radial pressure gradient along the full span of the suction side. The radial gradient favors the formation of a large clockwise-rotating structure, marked by the black arrows in the vorticity plots, as soon as the strut turning affects the flow ( $x_{mid} / C > 0.3$ ). It covers the whole duct and is called the vane passage vortex (VPV) (see the work of Spataro et al. [9]). The VPV deforms and shifts the vortices seen in the first plane ( $x_{mid} / C = - 0.01$ ) away from the endwalls. This movement can also be clearly seen in the next plane $x_{mid} / C = 0.73$ in Fig. 6. A strong pressure gradient from the pressure to the suction side can be seen in this plane ( $x_{mid} / C = 0.73$ ), which is caused by the strong turning of the main flow by the aft-loaded strut. The pressure gradient at the hub enhances the VPV and helps to transport fluid from the pressure toward the suction side of the strut. This displacement of fluid can be seen by the streamlines at the hub in Fig. 8b.

Looking now at the streamlines at the shroud (Fig. 7b), the pressure-side location of the stagnation point and the formation of the pressure-side and suction-side legs of the horseshoe vortex are clearly seen. In the planes $x_{mid} / C = - 0.01$ and $x_{mid} / C = 0.26$ of Fig. 6, the zones of positive and negative vorticities at the shroud are caused by the horseshoe vortices. The streamlines in Fig. 7b show that the suction-side leg diminishes when the strong turning of the flow starts. The strong resulting blade-to-blade pressure gradient provokes the evolution of an upper passage vortex that coalesces with the pressure-side leg of the horseshoe vortex. It is also seen in the large zone of counterclockwise rotating vorticity there (structure D in Fig. 6). This strong secondary flow is also visible in the loss map where the highest loss is caused by structure D. The other loss cores still stem from the HP stage and are deformed mainly by the VPV. Because of the sharp edge between the strut and the endwall, the vane passage vortex induces a corner vortex (CV) between the hub and the pressure side of the strut (structure CV).

The effect of the secondary flows discussed can also be seen in the streamlines at the suction and pressure sides of the strut. Figure 7a shows the pressure side where the splitters are slightly shifted for a better view to the strut rear part. The leading-edge vortices are forced up- and downward by the pressure gradient caused by the sweep design. Moving downstream along the pressure side, the streamlines in the upper part converge to a separation line $S_{1}$ indicating the position where the TLV (structure C) and the UPV (structure B) merged with structure E1 move away from the strut surface (compare plane $x_{mid} / C = 0.26$ with $x_{mid} / C = 0.73$ ). This is indicated by the arrows in Fig. 7a.

A second separation line $S_{2}$ is visible in the lower part of the strut, which is caused by the interaction of the VPV and the corner vortex. Both separation lines where two counter-rotating vortices meet are also marked in Fig. 6.

In Fig. 8b, the streamlines at the suction side of the strut are illustrated. The radial displacement at the hub is caused by the VPV, whereas the distinct radial shift at the shroud region is caused by the strong UPV (structure D).

Figure 9 shows the flow evolution in three selected planes for the baseline case. In the plane $x_{mid} / C = 0.73$ , the vortex structure is very similar to the embedded design, only TLV C is remarkably more pronounced (see Fig. 6). On the other hand, the pressure map of the baseline design shows a larger pressure gradient between the pressure and suction sides; the very low pressure at the suction side especially indicates a stronger flow acceleration there. This is probably caused by the missing blockage of the splitters. The higher pressure gradient of the baseline design increases the strength of the VPV, leading to higher losses caused by a stronger structure D, as visible in the loss map.

Fig. 9
Plots of $ω_{SW}$ , $C_{P}$ , and $ζ / \bar{ζ_{E}}$ at selected planes of the baseline design.

2. Splitter Channel

Plane $x_{mid} / C = 0.885$ in Fig. 6 is situated upstream close to the leading edge of the splitters and shows the first influences of the splitters on the flow. In the pressure distribution zones of high pressure, just in front of the splitters, can be seen in the upper part of the channel. The marked zone of high pressure ${FS}_{1}$ of splitter B is caused by a negative incidence resulting in flow separation (FS). The separated flow leads to a high loss, which is depicted in the loss map at plane $x_{mid} / C = 0.905$ in Fig. 6. The separated flow is also indicated by a small vortex, which can be observed in the $ω_{SW}$ map at plane $x_{mid} / C = 0.905$ .

In plane $x_{mid} / C = 0.885$ close to midspan, two smaller zones of higher pressure can be found at splitter B as well as at splitter A. The reason is again a negative incidence but, this time, the flow does not separate.

The maps of the streamwise vorticity in the planes $x_{mid} / C = 0.905$ and $x_{mid} / C = 0.98$ show that the secondary flow structures are chopped by the splitters and are not moved by the VPV anymore. The rotation of the large vane passage vortex is stopped by the splitters. The pressure distribution at plane $x_{mid} / C = 0.905$ shows nearly uniform pressure in the first two flow channels (I and II). Otherwise, the strong vorticity of the UPV (structure D) is still active between the splitters and causes a flow from the pressure side to the neighboring suction side in all three channels, as shown by the streamlines in Fig. 7b. The large UPV is split into three individual upper passage vortices in the three channels (UPV2 and UPV3). UPV3 moves toward the splitter B suction side and enhances the losses there resulting from the flow separation ${FS}_{1}$ . The contour plots show the highest velocity in channel I between the strut suction side and splitter A, whereas the lowest velocity is seen in the third channel between splitter B and the strut pressure side.

Looking at the baseline design at $x_{mid} / C = 0.905$ (see Fig. 9), similar vortex structures can be seen. The low-pressure region extends more to the channel midpitch because the flow is not confined by splitters. Again, the loss due to structure D is higher when compared to the embedded design.

The pressure distribution close to the channel exit (plane $x_{mid} / C = 0.98$ in Fig. 6) shows a reduction of the flow speed in the first channel, especially in the upper part, caused by an increase of the flow area. On the other hand, the available flow area of the second channel reduces, thus increasing the flow speed and reducing the pressure there. The third channel also shows a very low pressure, and thus high velocities on the splitter B suction side. The velocities are highest close to the hub and decrease in the radial direction due to the radial equilibrium.

In the streamwise vorticity plot in plane $x_{mid} / C = 0.98$ , a clockwise-rotating vortex between the two counterclockwise-rotating vortices of channel III (UPV3 and separation vortex ${FS}_{1}$ ) can be detected, which is formed as a buffer between the two corotating vortices. It is so strong that it generates the highest losses in the corresponding channel. The loss map also shows loss layers on the suction sides of all blades. At the splitters, they are caused by an enlargement of the boundary layer, which is a result of the incidence discussed previously. The loss on the strut suction side stems from the secondary flow structures coming from upstream (structure D).

In the hub corner of the strut suction side, a small structure ( ${FS}_{2}$ ) appears in plane $x_{mid} / C = 0.98$ . It is the result of a small flow separation there, which is also illustrated by the streamlines along the strut in Fig. 8b. Obviously, this separation also generates a loss that is visible in the loss map. Looking at the streamlines of the baseline design in Fig. 8a, a remarkably larger separation zone occurs in the strut trailing-edge region. The embedded design reduces this ${FS}_{2}$ structure, and thus the losses in this part of the channel, significantly, as was also observed in the oilflow visualization in [11]. This is caused by the acceleration due to the splitters, which leads to a reduced streamwise pressure gradient; therefore, the separation bubble gets smaller.

The effect of the splitters is studied by a comparison of the flow at the exit plane $x_{mid} / C = 1.02$ between the baseline (Fig. 9) and the embedded design (Fig. 6). The vorticity caused by the flow separation ${FS}_{2}$ is larger in the baseline case and induces a relatively strong counterclockwise-rotating structure. The corresponding streamlines (see Fig. 8) show a liftoff of the separated flow in the baseline design, leading to much higher losses in this region, which are clearly seen by a comparison of the streamwise vorticities of Figs. 6 and 9.

The overall distribution of the streamwise vorticity differs between the two designs in plane $x_{mid} / C = 1.02$ . Whereas the baseline design still shows the clear secondary flow structures coming from the upstream HP rotor and from the strut flow as already explained, the splitters chop the structures and the VPV does not exist anymore. These differences lead to more fluctuations in circumferential direction, but of lower amplitude, which will be discussed in the following.

The loss map of the baseline design (Fig. 9) shows a thick zone of high loss in the strut wake region, whereas in the embedded design, there are three smaller zones of high loss in the course of the wakes. There seems to be a reallocation of losses from the strut wake to the splitter wakes. A comparison of the loss cores of the strut wakes between the two setups shows that the nondimensional losses are, on average, approximately 50% lower for the embedded design. The same analysis at the core of the splitter wakes shows nearly doubled losses when compared to the same position of the baseline case.

C. Flow Downstream of the TMTF

A main goal of the embedded design discussed in this work is to homogenize the flow in order to achieve a more uniform flow entering the low-pressure rotor. To illustrate the advantages and improvements of the embedded design, the flow is compared with the baseline design at plane E between the TMTF and the LP rotor ( $x / C = 1.18$ ). Figures 10a and 10c show contour plots of the yaw angle $α$ , the normalized Mach number ${M a / M a}_{E}$ , the static pressure coefficient $C_{P}$ , and the normalized nondimensional loss ${ζ / ζ}_{E}$ ; whereas Fig. 10b gives a comparison of the radial distribution of the circumferentially averaged values. The Mach number and $ζ$ are normalized with their averaged values at plane E $M a_{E}$ and $ζ_{E}$ . The radial loss distribution is not normalized in order to allow a better comparison between the two designs. In this case, $C_{P}$ is defined as

C_{P} = \frac{p - \bar{p_{E}}}{\bar{p_{t, E}} - \bar{p_{E}}}

(5)

Fig. 10
Contour plots at plane E ( $x / C = 1.18$ ), downstream of the TMTF.

When comparing the contour plots, it is noticeable that the wakes of the struts are clearly visible in all diagrams for the baseline design and hardly recognizable for the embedded design. Also, the vane passage vortex can be detected in the baseline design, and especially well in the loss diagram (see also [9]). The spots of highest losses close to the hub and shroud in the wake region of the loss plot may be caused by corotating shed vortices between two neighboring VPVs at plane E, which are also visible in Fig. 9. At the hub, there is a superposition with the separation zone as shown by the loss contours. In contrast, in the embedded design, the shed vortices and the separation zone are still there but show smaller deviations compared to the averaged value. However, in the radial distribution, the embedded design shows higher losses beside the shroud region. This difference will be discussed in the following.

Comparing the yaw angle plots, there are stronger variations in the baseline case, leading to larger differences in the radial distribution. The stronger variations are caused partly by the VPV because it leads to an underturning (less swirl) close to the shroud and an overturning (higher swirl) of the flow close to the hub, respectively. Whereas the radial distribution of the inlet angle to the succeeding LP rotor can be considered in the LP blade design, a strong circumferential variation of the flow angle will lead to higher losses. Figure 11 therefore shows the circumferential variation of the yaw angle for three different spanwise positions. As mentioned previously, the embedded design leads to circumferential fluctuations of higher frequency but of lower amplitude. Looking at the standard deviation of the circumferential fluctuations at the three positions, we obtain 3.7, 5.0, and 2.5 deg for the baseline design and 2.8, 2.8, and 2.5 deg for the embedded design. Especially at midspan, the amplitude of the yaw angle fluctuations is about half of the value of the embedded design which will decrease circumferential inflow variations to the following rotor, and thus result in a positive effect on loss generation.

Fig. 11
Circumferential distributions of the yaw angle $α$ at different spanwise positions.

The Mach number distribution of the baseline design clearly shows the wakes of the struts and zones of highest velocity at the hub. Again, in the embedded design, there are less strong variations in circumferential direction (about half the standard deviation at midspan). The radial distribution is similar but showing more variations in the embedded design case. Similar conclusions can be drawn from the pressure distribution. The embedded design shows a more uniform pressure in the hub region; the effect of the different flow channels is hardly visible.

The loss map again shows more variations for the embedded design. The radial variations are quite similar, but the embedded design shows especially higher losses at the hub and lower losses at the shroud where the shed vortex is very active in the baseline design.

The most positive effect of the embedded design is a more uniform inflow to the succeeding LP rotor. On the other hand, the additional splitter blades can lead to higher losses in the TMTF. In Fig. 12, the mass-flow-averaged total pressure $p_{t}$ from the TMTF leading edge ( $x / C = 0$ ) to the trailing edge ( $x / C = 1$ ) is shown over the relative axial coordinate $x / C$ . The value of both simulations is normalized with their mass-flow-averaged total pressures $p_{t, LE}$ at the leading edge of the strut. It can be seen that the baseline design shows a higher decrease at the beginning compared to the embedded design. Whereas the total pressure of the baseline design continuously drops, the loss of the embedded design strongly increases where the flow is influenced by the splitters. This increase despite a reduced flow separation can be explained by higher flow velocities due to the decreased cross section and an increase of the wetted area due to the additional blades. This can be partly avoided by an adaption of the meridional flow section bringing the flow velocities down to the level of the baseline design. To what extent the higher losses can be compensated by an improved efficiency of the succeeding rotor due to the more uniform inlet flow has to be investigated in further studies.

Fig. 12
Comparison between the two setups of the mass-flow-averaged total pressure $p_{t}$ normalized to the mass-flow-averaged total pressure values at the leading edge $p_{t, LE}$ .

A further positive effect of the embedded design is a better guiding of the flow, which would allow us to reduce the number of strut vanes and splitter blades in a future design in order to retain the total weight of the TMTF. In the present case, the larger flow turning leads to a shift of the LP rotor stagnation point to the pressure side, and thus to incidence losses. This can be avoided by a modification of the rotor design. On the other hand, the work output is increased.

IV. Conclusions

The paper described the flow through a turning midturbine frame with an innovative embedded design. The duct flow has been computed with a steady-state CFD simulation. Measurement data have been used as inlet conditions, and the strut blade loading has been used to verify the agreement between measurement and simulation. The influence of the splitter blades on the flow has been analyzed with special emphasis on secondary flow phenomena.

The flow through the TMTF is influenced by large secondary flow structures coming from the upstream high-pressure stage. Within the duct, an uncommon large vane passage vortex dominates the flow evolution in the baseline design and leads to flow inhomogeneities. In the embedded design, the splitter blades successfully chop this vane passage vortex and new, but smaller, structures are formed. Whereas the highest loss occurs in the strut wake in the baseline design, the losses are more evenly distributed in the embedded design. In a succeeding study of the unsteady flow, Spataro et al. [12] and Bader et al. [14] proved experimentally and numerically a positive effect. Both publications showed a damping of the unsteady fluctuation by the splitters.

A loss analysis of the TMTF flow for both designs shows a higher total pressure loss of the embedded design due to blockage by the additional splitter blades, but an efficiency improvement of the succeeding LP rotor flow can be expected by the more uniform flow in the circumferential direction. However, it should be kept in mind that the subsequent low-pressure rotor has to be adjusted to the increased flow turning of the embedded design for a new setup.

The new design discussed in this paper is a promising design for future engines to reduce weight and rotor–rotor interactions. The present work can be considered as a first attempt to implement splitter blades in an already existing TMTF design, and its value lies in a thorough experimental and computational study on the losses and benefits of such a design. However, designers should consider the challenges emerging from the complex secondary flow structures to be expected so that a more uniform outlet flow can be obtained at similar or even lower losses.

C. S. TanAssociate Editor

Acknowledgments

The research leading to these results has been partially funded by the European Union (EU) within the EU DREAM project (contract no. ACP7-GA-2008-211861) as well as from the Austrian Federal Ministry for Finance. The authors would also like to thank the Austrian Federal Ministry for Transport, Innovation and Technology who funded the project ReLam within the Austrian Aeronautics Program TAKE OFF.

References

[1] Dominy R. G., Kirkham D. A. and Smith A. D., “Flow Development Through Inter-Turbine Diffusers,” Journal of Turbomachinery, Vol. 120, No. 2, 1998, pp. 298–304. doi:https://doi.org/10.1115/1.2841406 Crossref Google Scholar
[2] Norris G., Dominy R. G. and Smith A. D., “Flow Instabilities Within a Diffusing Annular S-Shaped Duct,” Proceedings of the ASME 1999 International Gas Turbine and Aeroengine Congress and Exhibition, Indianapolis, IN, June 1999. doi:https://doi.org/10.1115/99-GT-07 Google Scholar
[3] Wallin F., Arroyo Osso C. and Johansson G., “Experimental and Numerical Investigation of an Aggressive Intermediate Turbine Duct: Part I—Flowfield at Design Inlet Conditions,” 26th AIAA Applied Aerodynamics Conference, AIAA Paper 2008-7056, 2008. doi:https://doi.org/10.2514/6.2008-7055 Link Google Scholar
[4] Arroyo Osso C., Wallin F. and Johansson G., “Experimental and Numerical Investigation of an Aggressive Intermediate Turbine Duct: Part II—Flowfield Under Off-Design Inlet Conditions,” 26th AIAA Applied Aerodynamics Conference, AIAA Paper 2008-7055, 2008. doi:https://doi.org/10.2514/6.2008-7056 Link Google Scholar
[5] Marn A., “On the Aerodynamics of Aggressive Intermediate Turbine Ducts for Competitive and Environmentally Friendly Jet Engines,” Ph.D. Thesis, Inst. for Thermal Turbomachinery and Machine Dynamics, Graz Univ. of Technology, Graz, Austria, 2008. Google Scholar
[6] Göttlich E., “Research on the Aerodynamics of Intermediate Turbine Diffusers,” Progress in Aerospace Sciences, Vol. 47, No. 4, 2011, pp. 249–279. doi:https://doi.org/10.1016/j.paerosci.2011.01.002 PAESD6 0376-0421 Crossref Google Scholar
[7] Santner C., Paradiso B., Malzacher F., Hoeger M., Hubinka J. and Göttlich E., “Evolution of the Flow Through a Turning Mid Turbine Frame Applied Between a Transonic HP Turbine Stage and a Counter-Rotating LP Turbine,” Proceedings of the 9th European Turbomachinery Conference, European Turbomachinery Soc., Florence, Italy, March 2011, Paper 110. Google Scholar
[8] Lengani D., Santner C., Spataro R., Paradiso B. and Göttlich E., “Experimental Investigation of the Unsteady Flow Field Downstream of a Counter-Rotating Two-Spool Turbine Rig,” Proceedings of the ASME Turbo Expo, American Soc. of Mechanical Engineers, New York, 2012. doi:https://doi.org/10.1115/GT2012-68583 Google Scholar
[9] Spataro R., Santner C., Lengani D. and Göttlich E., “On the Flow Evolution Through a LP Turbine with Wide-Chord Vanes in an S-Shaped Channel,” Proceedings of the ASME 2012 Turbo EXPO, American Soc. of Mechanical Engineers, New York, June 2012.doi:https://doi.org/10.1115/GT2012-68178 Google Scholar
[10] Lavagnoli S., Yasa T., Paniagua G., Castillon L. and Duni S., “Aerodynamic Analysis of an Innovative Low Pressure Vane Placed in an S-Shaped Duct,” Journal of Turbomachinery, Vol. 134, No. 1, 2011, Paper 011013. doi:https://doi.org/10.1115/1.4003241 Crossref Google Scholar
[11] Spataro R., Göttlich E., Lengani D., Faustmann C. and Heitmeir F., “Development of a Turning Mid Turbine Frame with Embedded Design—Part I: Design and Steady Measurements,” Journal of Turbomachinery, Vol. 136, No. 7, 2013, Paper 071012. doi:https://doi.org/10.1115/1.4025950 Google Scholar
[12] Spataro R., Göttlich E., Lengani D., Faustmann C. and Heitmeir F., “Development of a Turning Mid Turbine Frame with Embedded Design—Part II: Unsteady Measurements,” Journal of Turbomachinery, Vol. 136, No. 7, 2013, Paper 071008. doi:https://doi.org/10.1115/1.4025949 Google Scholar
[13] Bader P., Sanz W., Spataro R. and Göttlich E., “Flow Evolution Through a Turning Mid Turbine Frame with Embedded Design,” Proceedings of the ASME 2014 Turbo EXPO, American Soc. of Mechanical Engineers, New York, 2014. doi:https://doi.org/10.1115/GT2014-25009 Google Scholar
[14] Bader P., Sanz W. and Spataro R., “Unsteady CFD Simulation of a Turning Mid Turbine Frame with Embedded Design,” Proceedings of the 11th European Turbomachinery Conference 2015, European Turbomachinery Soc., Florence, Italy, 2015, Paper 007. Google Scholar
[15] Hubinka J., Santner C., Paradiso B., Malzacher F. and Göttlich E., “Design and Construction of a Two Shaft Test Turbine for Investigation of Mid Turbine Frame Flows,” Proceedings of the ISABE 2009, ISABE Paper 2009-1293, International Association of Airbreathing Engines, Bedfordshire, U.K., 2009. Google Scholar
[16] Hubinka J., Paradiso B., Santner C., Pirker H. P. and Göttlich E., “Design and Operation of a Two Spool High Pressure Test Turbine Facility,” Proceeding of the 9th ETC Conference, European Turbomachinery Soc., Florence, Italy, 2011, pp. 1531–1540. Google Scholar
[17] ANSYS CFX-Solver Modeling Guide, Vol. 14.0, ANSYS, Inc., Canonsburg, PA, 2010, p. 416. Google Scholar
[18] Menter F. R., “Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications,” AIAA Journal, Vol. 32, No. 8, 1994, pp. 1598–1605. doi:https://doi.org/10.2514/3.12149 AIAJAH 0001-1452 Link Google Scholar

Tables

**Table 1 Thermodynamic properties of the used ideal gas**
Parameter	Value
Specific heat capacity	$c_{p} = 1.044 \cdot 10^{3} J / (kg \cdot K)$
Dynamic viscosity	$μ = 1.831 \cdot 10^{- 5} kg / (m \cdot s)$
Thermal conductivity	$λ = 2.61 \cdot 10^{- 2} W / (m \cdot K)$

Journal of Propulsion and Power, volume 33 issue 6 cover

Volume 33, Number 6November 2017

Metrics

Crossmark

Information

Copyright © 2017 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved. All requests for copying and permission to reprint should be submitted to CCC at www.copyright.com; employ the ISSN 0748-4658 (print) or 1533-3876 (online) to initiate your request. See also AIAA Rights and Permissions www.aiaa.org/randp.

Topics

Keywords

Acknowledgments

Digital

Received16 March 2016
Accepted17 March 2017
Published online30 May 2017

Flow Evolution Through a Turning Midturbine Frame with Embedded Design

Abstract