Structural and Multidisciplinary Optimization

, Volume 44, Issue 1, pp 125–136

Aero-mechanical optimisation of a structural fan outlet guide vane

Authors

    • Rolls-Royce Deutschland Ltd & Co KG
Industrial Application

DOI: 10.1007/s00158-010-0617-4

Cite this article as:
Clemen, C. Struct Multidisc Optim (2011) 44: 125. doi:10.1007/s00158-010-0617-4

Abstract

The aero-mechanical optimisation of a fan outlet guide vane (OGV) in the bypass duct of a modern turbofan engine is presented. The purpose of the described outlet guide vane arrangement is to remove the swirl from the flow coming from the fan and to connect the engine core structurally with the bypass duct and the engine mounts respectively. For that reason the outlet guide vanes have to fulfill aerodynamic requirements—such as low pressure loss and large working range and turning of the absolute flow to 0 degree—as well as the structural requirement to withstand the engine loads in all operating conditions. Such an arrangement has the advantage, that additional struts downstream of the fan outlet guide vane become obsolete, which is beneficial for the engine length and weight and hence the engine specific fuel consumption. At the same time the structural and aerodynamic requirements on such an outlet guide vane are intuitively contradictory. Therefore the design is more complex than for a conventional purely aerodynamic guide vane. To achieve the different requirements the conventional iterative design process has been replaced by a multi-disciplinary approach, which delivers an aero-mechanically optimised fan outlet guide vane geometry based on aerodynamic and mechanical boundary conditions and constraints. The prescribed boundary conditions are the structural load cases (flight conditions) and the aerodynamic inlet and outlet conditions for the fan design point. In addition an existing bypass duct geometry is used to get a valid comparison with a conventionally designed outlet guide vane/strut arrangement. In the structural optimisation, carried out as a parameter study, the effect of several geometrical parameters is investigated using the software ABAQUS. The aerodynamic optimisation is performed using the 2D-CFD-solver MISES on individual profile sections such that a minimum pressure loss and a maximum working range is achieved. For that purpose the profile shape is modified freely taking into account the constraints from the structural optimisation. A 3D-RANS-CFD analysis of the optimised vane and the comparison with the conventional vane/strut arrangement confirmed the improved performance of the chosen design approach and showed a significant reduction in pressure loss (ΔP/D) of nearly 20% compared to the conventional OGV/strut arrangement, leading to an SFC reduction of about 0.5%.

Keywords

Outlet guide vaneFanBypass duct

Nomenclature

\(\upbeta _{2}\)

Outlet angle

CFD

Computational fluid dynamics

D

Dynamic pressure

DoE

Design of experiment

Δ

Difference

ff

Fitness function

mm

Millimeter

MTO

Maximum take-off

MRT

Maximum reverse thrust

N

Newton

P

Total pressure

RANS

Reynolds-Averaged-Navier-Stokes

(S)OGV

(Structural) outlet guide vane

SFC

Specific fuel consumption

WEM

Whole engine model

\(\upomega\)

Loss

2D/3D

Two/three-dimensional

1 Introduction

Aiming at more environmental-friendly, more efficient and more powerful aero engines, structural outlet guide vanes (SOGVs) as replacement of plain (purely aerodynamic without structural task (Hughes 2001)) outlet guide vanes become more and more important in the further development of aero engines since this allows the deletion of the heavy bypass duct struts, which results in reduced masses, lower number of components to assemble, improved aerodynamic performance (increased efficiency factors) but also new load pathes compared to a conventional bypass duct design.

The present paper describes the aero-mechanical optimisation of a fan outlet guide vane (OGV) in the bypass duct of a modern turbofan engine. The purpose of the presented outlet guide vane arrangement is to remove the swirl from the flow coming from the fan and to connect the engine core structurally with the bypass duct and the engine mounts respectively. For that reason the outlet guide vanes have to fulfill aerodynamic requirements—such as low pressure loss and large working range and turning of the absolute flow to 0 degree—as well as the structural requirement to withstand the engine loads in all operating conditions.

Such an arrangement has the advantage, that additional struts downstream of the fan outlet guide vane, as described for example in Holewa et al. (2009), are not necessary anymore, which is beneficial for the engine length and engine weight and hence the engine specific fuel consumption (SFC). At the same time the structural and aerodynamic requirements on such an outlet guide vane are intuitively contradictory. Therefore the design is more complicated than for a conventional purely aerodynamic guide vane.

To achieve the different requirements not the conventional iterative design process has been chosen but a multi-disciplinary approach has been used, which delivers an aero-mechanically optimised fan outlet guide vane geometry based on aerodynamic and mechanical boundary conditions and constraints. Figure 1 shows the process chain starting with the structural optimisation, afterwards the aerodynamic optimisation is carried out, followed by a structural check of the new design, if the design passes this test it is ready for testing otherwise another iteration needs to be performed. The prescribed boundary conditions are the structural load cases (flight conditions) and the aerodynamic inlet and outlet conditions for the fan design point. In addition an existing bypass duct geometry is used to get a valid comparison with a conventionally designed outlet guide vane/strut arrangement. Furthermore an aerodynamically optimised leading edge shape (sweep) is defined (Clemen and Stark 2003).
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig1_HTML.gif
Fig. 1

Process chain

2 Structural optimisation/parameter study

The first step in the design process of the structural OGV (Fig. 1) is the structural optimisation of the OGVs (Fig. 2). Based on an extended front assembly finite-element-model without any structural OGVs, derived from a representative Whole Engine Model of an under-wing mounted engine, the Python scripting language was combined with ABAQUS CAE v6.81 for an automatic creation of a full parametric front assembly model including different OGV design parameters: number of OGVs, OGV material, thickness of OGV profile, top and bottom OGV chord lengths, top and bottom axial OGV offsets, radial sweep of leading edge (Clemen and Stark 2003), radial sweeping of trailing edge (Gustavsson 2006).
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig2_HTML.gif
Fig. 2

Structural optimisation process chain

The leading edge sweep is prescribed by a parametric definition shown in Fig. 3 and given in (1) (both as an example for the vane tip end only) to get an optimum aerodynamic performance and low noise emissions by an optimised combination of sweep angle and radial extent of the sweep along the blade height (Clemen and Stark 2003). The leading edge is swept in radial direction using an e-function to achieve a smooth shape. For that purpose the hub and tip sweep angles as well as the axial offset of the hub and tip leading edge points relative to the vane mid height axial position are design parameters. For more details see (Clemen and Stark 2003).
$$ \begin{array}{rll} \label{eq1} &&\mbox{Ax}\_{\rm Coord} \left[ {\% \text{height}} \right]\notag\\ &&{\kern10pt} = \frac{1}{5}\text{Rad}\_{\rm Extent}\_{\rm Sweep}\left[ {\% \text{height}} \right] \notag\\ &&{\kern16pt} \times \tan {\rm Sweep}\_{\rm angle}\_{\rm tip}\left( {1-e^{\frac{-5(100\% - \text{blade}\_\text{height}\left[ \% \right])}{\text{Rad}\_\text{Extent}\_\text{Sweep}\left[ {\% \text{height}} \right]}}} \right) \end{array} $$
(1)
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig3_HTML.gif
Fig. 3

Definition of leading edge sweep

This full parametric front assembly model is used for the structural optimisation of the OGVs by varying the different design parameters (number off vanes on the circumference, axial position of the vane in the bypass duct relative to the fan, vane leading edge shape/sweep, profile chord length, profile thickness, etc.) for the three relevant load cases: maximum reverse thrust, maximum take-off and hard landing (6 g down) such that minimum stresses, and deformation are achieved for all three cases.

At the same time the aerodynamic and acoustic requirements on vane count, vane thickness, leading edge sweep and position are taken into account in the structural optimisation process to rate the different design solutions.

Following conclusions can be drawn from the results of the structural optimisation study:
  • Figure 4 shows that the total number of the structural OGVs (investigated in the range of 35 to 48) around the front assembly circumference has only a slight influence on the resulting maximum von-Mises stress. Due to this the OGV number can be decreased down to 37/38 without any significant stress magnification, since an OGV number of 42 is preferred by the acoustics (cut-off design with 20 fan blades) this is chosen for the final geometry. In Figs. 4, 5 and 6 for comparison the stresses of a non-structural baseline OGV geometry with 44 vanes is depicted.

  • From the aerodynamic/aero-acoustic point of view the distance between the fan blades and the structural OGVs (SOGVs) has to be maximised to minimise the fan/OGV interaction. In terms of structural optimisation (Fig. 5) both the most forward and most rearward position of the OGV cascade relative to the bypass duct splitter nose at 0% are suitable designs, while the most rearward position (>50%) has the lowest absolute stress level and is more robust than the most forward position (0–10%) in terms of stress gradient with respect to position variation. Only an OGV position at the axial middle of the front assembly results in high OGV stresses. Therefore the rear OGV position has been chosen.

  • Radial sweeping of the OGV leading edge increases the maximum stresses. A nearly linear context between leading edge middle offset and stress increase could be determined. For that reason as a compromise only a small amount of sweep is introduced near hub and tip to achieve acceptable aerodynamic, acoustic and structural behavior.

  • The thinner the blades the higher the OGV stresses (Fig. 6), but only blades which are thinner than the initial thickness (t < 100%) result in a significant increase of OGV stresses. Blades thicker than the initial thickness (t > 100%) result only in slightly decrease of maximum vane stresses. With the maximum allowable von-Mises stress for the OGV structure of 90 N/mm2 (including a safety factor of 2.5) the minimum OGV chord length and minimum OGV thickness can be determined for all design variants and all relevant load cases. The resulting optimum thickness/chord combination for hub and tip is depicted in Fig. 7, whereas the thickness is a result of the structural parameter study and the optimum chord length will be a result of the aerodynamic optimisation.

  • The blade profile itself is not modified in the current study, the profile shape of the baseline OGV (controlled diffusion airfoil) is maintained. The profile shape is optimised in the aerodynamics part, see Section 3.

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig4_HTML.gif
Fig. 4

Effect of OGV vane count

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig5_HTML.gif
Fig. 5

Effect of axial position

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig6_HTML.gif
Fig. 6

Effect of thickness laser operating system

Summarising all results, generally the design of a structural OGV for the given assembly has to include the following characteristics:
  • rear (downstream) OGV position with reference to the casing,

  • bottom OGV length has to be maximised,

  • number of OGVs should be higher than 37,

  • no need for significant increase of OGV thickness.

Combined with aerodynamic (swept leading edge) and acoustic (number of vanes) requirements a vane with an aerodynamic leading edge sweep as defined by (1) and 42 vanes at the most axially rearward position, as shown in Fig. 5 and a radial chord length distribution as shown in Fig. 8 has been chosen, the related hub and tip thickness is shown in Fig. 7.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig7_HTML.gif
Fig. 7

Optimal combination of thickness and chord

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig8_HTML.gif
Fig. 8

Optimised SOGV

Based on the above described parameter settings, the aerodynamic optimisation of the structural OGV (SOGV) is carried out. This optimisation is described in the next chapter. The outcome of the aerodynamic optimisation is the final geometry, which has been has been integrated into the original front assembly model to verify the new design.

The result of that analysis is, that the resulting maximum blade stresses are within the stress limit of 90 N/mm2 for all three load cases, which means that the aerodynamic optimisation fulfilled the given constraints.

3 Aerodynamic optimisation

The aerodynamic optimisation is carried out at four two-dimensional profile sections of the SOGV along the duct height: 10%, 30%, 50% and 90% as shown in Fig. 9. At each section three different operating points are included: the design point and two off design points. The optimisation tool used is the DLR Auto Opti (Voss and Nicke 2008), the flow field is calculated with MISES (Drela and Youngren 1996). At the end the four optimised profiles are combined with the prescribed blade parameters from the structural optimisation to a three dimensional blade surface.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig9_HTML.gif
Fig. 9

Two dimensional profile sections

The optimisation objectives are to minimise the total pressure loss (\(\upomega _{0})\) and to minimise the deviation from the intended outlet flow angle (\(\upbeta _{2})\) of 0 degree. With the appropriate fitness functions ff1 and ff2 implemented as follows:
$$ \label{eq2} \mbox{f\/f1} = 0.50 \upomega_{\rm {O}\_P0} +0.25 \upomega _{\rm O\_P1} + 0.25\upomega _{\rm O\_P2}, \mbox{f\/f2} = \upbeta _{2\_{\rm P0}} $$
(2)

The total pressure loss is weighted with 50% at the design point (P0) and with 25% at the two off design points (P1 and P2). The outlet flow angle is accounted for only at the design point (P0).

The aerodynamic optimisation process chain can be described with the following steps:
  1. 1.

    Generation of members. Members are sets of optimisation parameters. In this optimisation all 11 optimisation parameters are part of a parameter set that specifies the shape of the SOGV profile. The parameters are: stagger angle, leading and trailing edge angle, four coordinates of control points defining the suction side, position of the maximal thickness, two parameters defining the shape of the thickness distribution, axial length. The first 3000 parameter configurations are generated with the “Latin Hyper Cube” method (Design of Experiment). This method generates random members in a way that the members are spread throughout the entire parameter space. By doing so there remain no large and unwatched parameter subspaces. The result of the DoE is used to define sensible limits for the different design parameters. Once the initial members have passed the process chain further members are generated by a genetic algorithm (Falkenauer 1994; Goldberg 1989). The strategies included are differential evolution, mutation and crossover.

     
  2. 2.

    Calculation of the maximum thickness parameter. This is an additional step inserted to the standard process chain. It takes care of the special geometric restriction to the SOGV given by the structural constraints. As a result of the structural investigations the maximal thickness is defined as function of the axial length derived from Fig. 7.

     
  3. 3.

    Generation of the SOGV profile. The DLR Blade Generator generates a MISES readable SOGV profile from the related parameter set.

     
  4. 4.

    Calculation of the flow field with MISES.

     
  5. 5.

    Calculation of the fitness values. The fitness values are calculated with the given fitness functions. In case the MISES calculation didn’t converge, the fitness values are set to some penalty values.

     
  6. 6.

    Storage of the member in the database, i.e. the optimisation parameters and the fitness values.

     
  7. 7.

    Ranking. All members in the database have an assigned rank that reflects the quality of its fitness values in comparison to all other members. Each time a new member is added to the database the rank of all members is updated. Members with a small rank will more likely be chosen by the genetic algorithm to be parents for the generation of new members.

     
The optimisation process for each duct height includes several optimisation runs with small adjustments of the parameter boundaries and fine tuning of the optimisation control parameters. Altogether about 20.000 up to 40.000 members are calculated. Since the time for one process iteration is only a couple of seconds, the whole optimisation has been carried out within some days. Additionally it turned out that the chosen number of runs was unnecessarily high, so for future application it can be reduced significantly. Figure 10 shows the fitness values of the best 10.000 members for the 50% section and the members with rank 1 (Pareto front). The member chosen is marked by “final profile” in Fig. 10 and is the member which delivers for 0 degree exit angle the lowest loss value.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig10_HTML.gif
Fig. 10

Best 10,000 members for mid-section (left), Pareto front (right)

The aerodynamic optimisation has been applied at first to the mid-section. Coming from the resulting axial length for the mid-section, the axial length for the other duct heights (10, 30 and 90%) is set to match the optimised axial length of the mid-section and the leading edge sweep and trailing edge position defined by the structural optimisation, see Fig. 8. The absolute maximum thickness of the other sections is set to the mid-section value to achieve a constant maximum profile thickness over the whole duct height.

The aerodynamic boundary conditions for the MISES CFD are an inlet Mach number, an inlet static to total pressure ratio, a turbulence level and the inlet flow angle (varying with the three investigated operating points).

The free geometrical parameters are: profile stagger angle, profile leading and trailing edge angle, four coordinates of control points defining the suction side shape, position of the maximum thickness, two parameters defining the shape of the thickness distribution, axial length. Based on the axial length the maximum thickness is derived from Fig. 7.

Figure 11 show a comparison of the optimised mid-section profile and the baseline mid-section profile of the non-structural vane. In Fig. 12 the MISES results for the two profiles are depicted. The results clearly show, that the optimised profile has lower losses for the two off-design conditions, marked by the arrows in Fig. 12, and a much larger working range than the reference profile, only at the design point the losses are marginally higher.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig11_HTML.gif
Fig. 11

Optimised (green) mid-section versus reference profile (red)

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig12_HTML.gif
Fig. 12

Mises 2D-CFD results of optimised (red) mid-section versus reference profile (black)

In order to generate a 3D blade surface from the optimized profiles at 10%, 30%, 50% and 90% it is necessary to define the profiles at the inner and outer duct walls. This is done by copying the profile parameters from 10% to 0% and from 90% to 100% duct height. The axial length though is extended to match the given leading edge sweep. The only parameter that is adjusted for the extrapolated sections is the stagger angle. The profile shapes in between are generated by interpolation using the axial length and leading edge shape from the structural optimisation. This results in a smooth surface. Figure 13 shows the final blade surface.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig13_HTML.gif
Fig. 13

Final 3D blade model

4 CFD validation

The OGV geometry resulting from the structural and the aerodynamic optimisation with 2D CFD (MISES) is validated with 3D CFD calculations of the fan stage including spinner, fan, splitter, engine section stator, 42 SOGVs and bypass duct (Fig. 13) and is compared with a conventional setup with 44 purely aerodynamic OGVs and 10 downstream struts for the structural connection between core and bypass duct.

The grid used is structured with O- and H-mesh topologies and generated with the Rolls-Royce grid generator PADRAM (Shahpar and Lapworth 2003). The individual blade rows are meshed single passage with mixing planes between the adjacent rows. The grid has 2.7 million grid points. The calculations are performed with the Rolls-Royce CFD code Hydra (Lapworth 2004) in steady mode using the Spalart-Allmaras turbulence model (Spalart and Allmaras 1992).

The boundary conditions for the CFD calculation are total pressure, total temperature, whirl and radial angle profiles at the fan inlet for the stage calculation. At the exit of the bypass duct a static back pressure (radial equilibrium) is defined to achieve the appropriate fan running condition. The core is running on a fixed flow function.

A comparison of the fan performance for the conventional OGVs and for the SOGVs shows, that the fan is nearly not affected by the different sets of OGVs. The fan efficiency and stability are maintained compared to the case with OGVs and downstream struts.

The losses predicted by Hydra for the new SOGVs and the conventional setup of OGVs + struts are listed in Table 1. The losses are values related to the baseline OGV loss which is by definition 1 and the accounting planes to determine the losses are identical for the original and the optimised configuration. A significant reduction in pressure loss (ΔP/D) of nearly 20% compared to the conventional OGV/strut arrangement can be observed, leading in terms of ΔP/P to an SFC reduction of about 0.5%. The SOGV configuration (42 vanes) has lower losses than the baseline configuration consisting of 44 OGVs and 10 struts, while the SOGV compared to the OGVs only has slightly higher losses. This is also visible in the circumferentially averaged radial loss distribution in Fig. 14. The SOGV features less endwall loss at the casing, due to the leading edge sweep, but higher losses at mid height due to the increased chord and thickness. One reason for the overall lower loss is, that the interaction of OGVs and struts, which generates a significant upstream pressure effect, is avoided if the SOGVs are used. In that configuration the circumferential upstream static pressure field is nearly constant, see Fig. 15.
Table 1

Loss summary

 

SOGVs

OGV+Strut

 

ΔP/P

ΔP/D

ΔP/P

ΔP/D

OGV

1.09

1.18

1.00

1.00

Strut

0.44

0.56

Total

1.09

1.18

1.45

1.44

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig14_HTML.gif
Fig. 14

Circumferentially averaged radial loss distribution for SOGV (red) and OGV (black)

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig15_HTML.gif
Fig. 15

Circumferential static pressure distribution at mid-height upstream OGV leading edge

Figure 16 shows that the SOGVs deliver the same exit flow angle as the OGVs. Figure 17 shows the passage Mach number contours at mid height for OGV and SOGV. Due to the increased thickness the SOGVs have a slightly higher suction side peak Mach number and hence a locally lower suction side static pressure.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig16_HTML.gif
Fig. 16

Circumferentially averaged radial whirl angle distribution for SOGV (red) and OGV (black)

https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig17_HTML.gif
Fig. 17

Mach number contours at mid height for OGV (left) and SOGV (right)

The loss improvement for the SOGV is also confirmed by the absolute total pressure field downstream in the bypass duct in Fig. 18 for both configurations. While the SOGV configuration features thicker OGV wakes and slightly more secondary flows on the end-walls, the conventional configuration has overall higher losses due to the additional losses by the downstream struts, which generate significant wakes and endwall flow field disturbances.
https://static-content.springer.com/image/art%3A10.1007%2Fs00158-010-0617-4/MediaObjects/158_2010_617_Fig18_HTML.gif
Fig. 18

Total pressure wake downstream of OGV/Strut (left) and SOGV (right)

5 Summary

In the present paper the aero-mechanical optimisation of a fan outlet guide vane (OGV) in the bypass duct of a modern turbofan engine is presented. The purpose of outlet guide vane arrangement is to remove the swirl from the flow coming from the fan and to connect the engine core structurally with the bypass duct and the engine mounts respectively. For that reason the outlet guide vanes have to fulfill aerodynamic requirements as well as the structural requirement to withstand the engine loads in all operating conditions. To achieve the different requirements the conventional iterative design process has been replaced by a multi-disciplinary approach, which delivers an aero-mechanically optimised fan outlet guide vane geometry based on aerodynamic and mechanical boundary conditions and constraints. The structural analysis is performed with the software ABAQUS.

The profiles of the structurally optimised OGV are aerodynamically optimized at four two dimensional sections to produce 1) a minimal total pressure loss at the design point and two off design points and 2) a zero degree outlet flow angle at the design point. The two-dimensional MISES calculations promise that the total pressure loss of the structural OGV is approximately of the same size than that of the original conventional OGV.

A 3D-CFD analysis of the optimised vane with 3D RANS and the comparison with the conventional vane/strut arrangement confirmed the improved performance of the chosen design approach and showed a significant reduction in pressure loss (ΔP/D) of nearly 20% compared to the conventional OGV/strut arrangement, leading in terms of ΔP/P to an SFC reduction of about 0.5%.

The investigation proves that a configuration with structural OGVs will give similar or even lower losses than a conventional configuration with OGVs and struts and that the chosen multi-disciplinary optimisation approach is the right methodology to achieve the presented results.

Acknowledgements

The author would like to thank the government of Brandenburg for funding the presented work in the frame of the research project OPAL and the management of Rolls-Royce Deutschland Ltd. & Co. KG for the permission to publish this work. Thanks to Axel Holewa (DLR) and Dr. Thomas Klauke for their contribution to this work.

Copyright information

© Springer-Verlag 2011