Enhanced Gas Turbine

Performance Simulation Using

CFD Modules in a 2D
Representation of the
Friederike C. Mund Low-Pressure System for a
Georgios Doulgeris
High-Bypass Turbofan

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

Pericles Pilidis
School of Engineering,
The improvement of performance simulation for gas turbines has been approached in
very different ways. In particular for high-bypass turbofans, efforts have been made to
Cranfield University,
investigate radial flow distributions. The aim of the presented study was to combine a
Cranfield MK 43 0AL, UK
conventional characteristics-based performance code using a 2D representation of the
fan with 2D representations of the adjoining intake and bypass system. Computational
fluid dynamics (CFD) was the chosen tool to generate modules for the intake, bypass
duct, and bypass nozzle. This approach required geometry data. A design procedure to
generate these components in an axisymmetric meridional fashion and the numerical
requirements for the CFD modules were developed. Typical component performances
were predicted and the combined use of CFD and the performance code showed that in
terms of performance, the inclusion of intake and bypass losses and the radial inlet
distribution was worth considering. In particular, however, the required numerical effort
was significant. 关DOI: 10.1115/1.2364197兴

Introduction geometry data exchange. A stepwise path to achieving a higher

order of fidelity for modeling an engine is to replace individual
The three-dimensional numerical simulation of a gas turbine
engine components of a 0D performance simulation tool with 2D
engine as a whole is leading edge research. The scale of such a
modules. This reduces computational complexity significantly and
computation limits its application to case studies outside the stan-
depending on the number and type of 2D components, generally
dard design procedure of aeroengines. At the other end of the
can be performed on a single processor or personal computer.
scale, gas turbine performance simulation based on 0D component
The use of 2D modules recognizes the fact that radial inlet
characteristics is well integrated into the design process.
profiles influence the performance of the components. For 0D
For the past two decades the improvement of gas turbine per-
performance models, no extrapolation mechanisms are available
formance modeling by using higher fidelity component models
to derive radial profiles at component interfaces. This limitation
has been pursued. Leading the field is the NASA numerical pro-
can only be overcome by starting at the most upstream compo-
pulsion system simulation 共NPSS兲, a multidisciplinary environ-
nent, the intake, using a radial modeling approach and following
ment using state-of-the-art computing and communication tech-
the flow downstream.
nology for the analysis and design of aircraft engines. NPSS
A 0D performance model for a two-shaft, high-bypass turbofan
supports “numerical zooming” from 0D, 1D, 2D, and 3D compo-
with a radial representation of the fan was made available to the
nent engine codes in terms of aerodynamics, structural engineer-
authors. As the backbone of the performance simulation, it pro-
ing, and heat transfer analyses. In close collaboration with engine
vided radial boundary conditions downstream of the fan and with
manufacturers, 3D Navier–Stokes simulations of the low-pressure
some modifications was developed to be capable of handling ra-
system 共LPS兲 including external flow, nacelle, inlet, fan, bifur-
dial inlet profiles. The next logical step was to develop radial
cated bypass and core inlet, bypass vanes, core inlet guide vanes,
representations of the intake, bypass, and bypass nozzle. The cho-
mixer, and exhaust nozzle have been performed. These were
sen numerical tool was CFD, which can be implemented in 2D
coupled with the core components, modeled using a 0D perfor-
when assuming axisymmetry. Geometrical data, in this case the
mance simulation. To keep such complex models within the scope
meridional cross sections of the components, were required but, in
of industrial application, a solution turn-around time for a single
an academic environment, detailed engine data are generally un-
operating point was limited to “overnight solutions” and was fa-
available. The simplification to axisymmetric geometries reduced
cilitated by using parallel computing techniques using in excess of
the extend of the data required and enabled the data set to be
32 processors 关1兴.
generated using basic design principles.
This certainly represents the high end of the scale of current
performance modeling and was applied to specific engines in
close collaboration with engine manufacturers to facilitate test and
Fan Modeling
A component of particular importance for radial modeling is the
Contributed by the International Gas Turbine Institute 共IGTI兲 of ASME for pub- fan, especially for high-bypass turbofans when the hub-to-tip ratio
lication in the JOURNAL OF ENGINEERING FOR GAS TURBINES AND POWER. Manuscript
received October 1, 2005; final manuscript received February 1, 2006. IGTI Review
of the fan is relatively low and the flow parameters vary consid-
Chair: R. S. Abhari. Paper presented at the ASME Turbo Expo 2006: Land, Sea, and erably across the span. As the fan delivers compressed air to two
Air 共GT 2006兲, Barcelona, Spin, May 8–11, 2006, Paper No. GT2006-90440. downstream systems, the core and the bypass, the inlet pressures

Journal of Engineering for Gas Turbines and Power JULY 2007, Vol. 129 / 761
Copyright © 2007 by ASME
 Table 1 Performance data for the two-shaft high-bypass tur-
bofan at design point

Thrust 共N兲 20000 N

BPR 3.75
Mass flow 79.04 kg/ s
Overall pressure ratio 27.8

Fig. 1 Performance model for a two-shaft high-bypass

turbofan
adapted as a function of the radial inlet distribution of total pres-
sure. To preserve the 0D maps, the resulting pressure ratio was
and temperatures to these systems may differ considerably. The kept the same. Provided that the incidence angles were unaffected,
fan therefore provides a strong incentive for an extension to radial no changes in radial inlet velocities occurred, allowing for the
modeling. assumption of the same fan efficiencies. On this basis, the radial
Riegler et al. 关2兴 presented a fan model for a three-stage low-

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

distribution of the fan efficiency remained unchanged. A further
bypass turbofan where the radial flow profiles at the exit of the fan simplification was the assumption of unchanged radial distribu-
were derived from rig data. It was shown that the bypass ratio tions of the flow function. The overall reduction of engine inlet
influenced the fan map parameters and a correcting algorithm was mass flow due to radial distortion was accounted for by introduc-
presented accounting for this effect. Based on the assumption that ing the inlet pressure recovery factor 共PRF兲, defined as the ratio of
the splitter was located far downstream of the fan exit, the sug- total pressure at the fan-inlet to the free-stream stagnation pres-
gested approximation would deteriorate with closer proximity of sure of the captured flow, into the simulation.
the splitter to the fan exit. For each whole-engine simulation, a new set of profiled char-
Curnock et al. 关3兴 used a similar fan model for high-bypass acteristics was generated for the relevant inlet conditions. This
turbofan applications. The radial distribution of the pressure ratio, was facilitated by means of a separate Fortran program combined
efficiency, and mass flow of the fan were stored in “profiled” with a MATLAB procedure.
characteristics. Notional radial profiles were generated, not spe-
cific to a particular fan. However, the artificial profiles were con-
sidered to be realistic for a high-bypass turbofan. Previous inves- Design of a Generic Propulsion System
tigations by Marshall 关4兴 showed that fan profiles were CFD calculations require boundary conditions specified in
independent of the downstream flow split. Deviations in engine terms of flow parameters and geometrical boundaries. A CFD
performance of up to 2% in thrust, 1% in specific fuel consump- study therefore inevitably attributes dimensions to the traditional,
tion 共SFC兲, 3% in bypass ratio 共BPR兲 and 8% in core pressure nondimensional performance prediction. Performance simulation
ratio were found when using the 2D representation in comparison considers component losses and mass flow balances, but geo-
with values appropriate to the 0D fan characteristic. A further metrical areas can only be derived where additional assumptions
investigation of the influence of radial inlet distortion on the en- were made for the static pressures, as for the choked nozzle-exit
gine performance indicated an equivalent reduction of thrust and condition, for example. Moreover, data are only available at com-
bypass ratio for an intake pressure loss of 1%, along with an ponent interfaces, neglecting the streamwise dimension. Such data
equivalent increase of SFC 关5兴. sets are insufficient for performing two-dimensional CFD simula-
It was in the nature of both of these models that the fan maps tions and additional data have to be included.
utilized referred to a specific inlet condition. Riegler et al. 关2兴 When engine manufacturers base their new engine designs on
derived the maps from experimental data, which inevitably re- existing ones, first geometry data are usually in place for prelimi-
ferred to the test cell fan-inlet profile. Inlet distortion, however, nary investigations. In an academic environment, detailed geom-
affects the fan-exit profiles and a coupling of such fan models to etries are generally not available. The absence of this essential
an upstream component would require changes of the profiled information required the design of a generic geometry data set.
data and the maps. The general design procedure for the intake, nacelle, bypass, and
The streamline curvature method can handle radial inlet data, bypass nozzle is outlined in the following sections.
but so far it was generally used further into the engine design
process, for turbomachinery design and analysis. Recently, its use Subsonic Intake and Nacelle. The design process for the in-
for a multistage fan coupled with a 0D performance model has take and nacelle geometry compromised the following steps:
been presented for a low-bypass turbofan by Li 关6兴, with encour-
aging results. Higher benefits can be expected for single-stage • sizing of the intake
fans. An adaptation of the model for a high-bypass turbofan ap- • establishment of key internal intake design parameters
plication, however, has not been available to the authors and • design of the internal lip contour
therefore some adaptations were made to the radial fan model by • design of the diffusion section
Curnock et al. 关3兴. • establishment of key external nacelle design parameters
• design of the forebody contour
• layout of the afterbody shape
Performance Model of the High-Bypass Turbofan
The performance model by Curnock et al. 关3兴 with a 2D repre- The air intake limits the mass flow that can enter the engine.
sentation of the fan was used for the study. A schematic of the Therefore, the intake throat, as the smallest area, has to be large
performance model of the investigated two-shaft, high-bypass tur- enough to meet the maximum flow demand of the engine. The
bofan with separate exhausts is illustrated in Fig. 1. The shaded design procedure for the intake selects the throat Mach number
components were modeled two-dimensionally, whereas the re- and the downstream diffusion rate to provide the right fan-inlet
maining components employed 0D characteristics. A summary of velocity. A schematic of the dimensions of a subsonic intake and
the performance data at design point at the end of climb is given the surrounding engine nacelle is shown in Fig. 2. The key geo-
in Table 1. metric parameters used for intake design are the contraction ratio
The characteristics of the fan were based on undistorted inlet 共CR兲, defined as the ratio of highlight area to throat area and the
flow. Profiled characteristics were available for the fan flow func- gross diffusion ratio 共DRgr兲, defined as the ratio of the circular
tion, pressure ratio, and efficiency. To reflect radial changes of area defined by the fan tip radius to throat area. The net diffusion
inlet pressure in the fan-exit profile, the pressure ratio profile was ratio 共DRnet兲 is based on the net inlet annulus area of the fan,

762 / Vol. 129, JULY 2007 Transactions of the ASME

 Table 2 Design guidelines for civil high-bypass turbofan
intakes.

Design range

Fan hub-to-tip ratio r2,hub Ⲑ r2,tip 0.3 0.3. . . 0.6a

Contraction ratio CR 1.25 1.2. . . 1.35b
Diffusion ratio DRgr 1.125 1.25. . . 1.35b
Throat Mach No. 0.72 0.7. . . 0.75b
Diameter ratio DH Ⲑ Dmax 0.815 0.8. . . 0.9c
Fig. 2 Key design parameters for a propulsion system Boat-tail angle ␤ 16 deg ⬍16 degb
a
Sec Ref. 关8兴.
b
accounting for the presence of the spinner. Sec Ref. 关14兴.
c
It is crucial to size the intake according to the highest engine- Sec Ref. 关12兴.
flow demand to avoid choking of the intake under any condition

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

within the flight envelope. The nondimensional mass flow rises
with altitude and turbine entry temperature 共TET兲 and the sizing be weakened and directed towards the thickest section 关12兴.
point for the intake is usually set at the end of climb. The cruise The NACA 1-series cowl-forebody profiles are based on a pro-
altitude and aircraft cruise speed are just reached, but the TET is file that can be varied in length and diameter ratio. They have
well above the cruise value. Cumpsty 关7兴 suggested a TET of gained wide acceptance for practical use, it being crucial to use an
125 K below the take off condition for the end of climb. The appropriate length ratio for the profile. The intake critical mass
sizing point was therefore modeled at 10670 m altitude, a Mach flow ratio and drag-rise Mach number can be correlated against
number of 0.83, and a TET of 1575 K, and the predicted engine the forebody dimensions, based on the empirical relationships de-
mass flow and relevant atmospheric conditions were used to size veloped by Stanhope 关13兴. A drag-rise margin was introduced by
the fan. increasing the Mach number at cruise by 0.08 for the chosen
Based on the maximum throat Mach number, DRgr, and the fan layout of the forebody. The resulting forebody profile following
hub-to-tip ratio rhub / rtip, the key intake parameters were derived the NACA nomenclature can be designated as NACA-1-81-80.
using the “Q function” 共see Fig. 3兲. To allow for future flow The coordinates of the profile were based on the geometry rep-
requirements, a “flow stretch” of 5% was introduced. resentation of the NACA-1 series by ESDU 关14兴 for computa-
Fans generally have hub-to-tip ratios ranging from 0.3 to 0.6 tional fluid dynamics. This analytical approximation of the origi-
关8兴, where high-bypass turbofans tend to have lower values to nal data was designed with improved smoothness of the second
achieve the highest practical mass flow per unit frontal area 关9兴. derivatives to avoid irregularities in the pressure distributions in
The recommended design ranges for other key intake and nacelle flow computations.
parameters are summarized in Table 2. A lower DRgr was selected, The afterbody length was designed using the maximum nacelle
as comparable values were found for commercial engines of the diameter, which was based on the selected diameter ratio, together
same size. The net internal diffusion ratio was relatively small and with the boat-tail angle and the exit nozzle areas. The aerody-
a linear streamwise diffusion being chosen. With a generic spinner namic design of an axisymmetric afterbody involved fairing the
geometry from a similar engine, the internal inlet contour was boat tail between the maximum nacelle diameter with the final
designed. nozzle. A circular arc profile was used. Less curvature avoided
A typical CR of 1.25 was chosen. The entry lip, as one of the premature drag rise in the expansion and therefore a flow-suction
most important components of a subsonic inlet, connected the region just downstream of the maximum diameter. The boat-tail
highlight with the throat plane: A super-ellipse curve for improved angle ␤ was chosen such that boundary layer growth was limited
take-off performance was chosen with a ratio of a / b of 2.0 and an and flow separation avoided 关15兴.
elliptic exponent of 2.0 关10兴.
The selection of the nacelle forebody geometry is often a com- Bypass-Duct and Nozzle. The bypass-duct and nozzle designs
promise to satisfy conflicting requirements. The avoidance of were based on the area distribution between the fan exit and
wave drag at high Mach numbers requires a long, thin cowl, but bypass-nozzle exit. The bypass-nozzle radii were established with
good low-speed spillage-drag characteristics are associated with the design of the afterbody and the defined bypass-nozzle exit area
short, fat cowls 关11兴. An intake cowl creates local velocities on the for the performance model of the engine. As found in McKenzie
external surface that are in excess of the free-stream velocity. With 关9兴, the fan-exit tip radius was assumed to be 0.98 of the inlet
increasing aircraft speed, a region of supersonic flow will develop. radius. With recommended bypass-inlet Mach numbers between
This may also be accompanied by flow separation, leading to 0.3 and 0.4 by Walsh and Fletcher 关16兴, a fan-exit hub radius of
rapid drag rise. Keeping the surface flow velocities subsonic gen- 1.25 r2,hub resulted for cruise conditions. Downstream of the fan
erally requires diameter ratios of the highlight to the maximum exit, up to the position of the leading edge of the splitter, a con-
nacelle diameter DH / Dmax of around 0.5–0.7. For high-bypass tur- stant duct cross-sectional area was assumed.
bofans this ratio usually is in the region of 0.8–0.9. But with The area distribution between the fan exit and bypass nozzle is
careful control of the curvature the formation of a shockwave can illustrated in Fig. 4. As illustrated with the solid line labeled “De-
sign,” a constant area over 70% of the entire length of duct and
nozzle was assumed, followed by a converging section with a final
area reduction to 64% of the bypass-inlet area. The shape of the
converging area reduction was designed using a sixth order poly-
nomial function to achieve a smooth area-change transition be-
tween the constant and the converging section. At the nozzle exit,
the shape was mainly governed by the external shape of the
nacelle—the outer contour of the bypass having to stay within the
nacelle outline and a certain distance having to allow for manu-
facturing clearances. A decreasing slope of the area reduction was
necessary to achieve such a configuration.
The inner and outer contour of the bypass duct and nozzle were
Fig. 3 Procedure to derive key intake design parameters designed using polynomial functions. The outlines are shown in

Journal of Engineering for Gas Turbines and Power JULY 2007, Vol. 129 / 763
 intake lip. The main mesh in Fig. 5 used an H structure where the
gridlines led from the left to the right block boundary and the
orthogonal gridlines connected the top with the bottom block
boundaries. For air intakes operating at forward speed this pro-
vided an appropriate mesh structure as the resulting streamlines
are closely aligned with the gridlines.
In 共b兲, a thinner boundary layer region was chosen and the
edges of the boundary layer mesh 共BLE1 to BLE4兲 moved within
the overall mesh block. Without GAMBIT’s boundary layer func-
tion, a decomposition of the domain into subregions would have
been required to apply structured meshes. This implied that for
different boundary layer thicknesses, different block structures
would be necessary, adding complexity to the preprocessing pro-
Fig. 4 Streamwise distribution of the cross-sectional bypass cedure and reducing flexibility for general application.
area and bypass contours
The meshes illustrated in Fig. 5 were generated using GAMBIT’s

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

boundary layer mesh function, showing the independent align-
ment of boundary layer and main mesh. However, the feature did
the lower section of Fig. 4. The area distribution of a scaled draw-
not work reliably and despite using the latest release, close dis-
ing of a similar commercial engine was also included 共dashed line
cussions with the code vendors and intensive trials, the desired
labeled “Drawing”兲 to demonstrate how misleading such illustra-
meshes could not be achieved in all cases. Correct definition of an
tions can be if used for technical investigations.
independent alignment of boundary layer mesh and main mesh,
often still resulted in the boundary layer running parallel to the
Setup of 2D CFD Modules main mesh. A significant effort was necessary to establish semi-
Once a geometry is established, several issues have to be con- reliable procedures in order to achieve the correct mesh structures
sidered: for the most important cases. If a flawless mesh file for the CFD
modules was available, scaling and mapping may be used to cre-
• shape and size of the computational domain ate similar face meshes. Care had to be taken when remeshing
• grid structure and density edges. Solving the problem was beyond the scope of this study
• boundary conditions and and the code vendors were made aware of the problem. But for
• solver settings future activities involving structured meshes and boundary layers,
other mesh generators should be considered.
For the present case, a 2D axisymmetric setup with structured
meshes was selected. The structured approach was considered to CFD Module of the Intake. Initial trials were performed using
be appropriate as it allowed the mesh to be aligned with the flow a 1/4 circular shape of the computational domain, where when
in most regions of the domain and thus improved numerical ac- rotated around the symmetry axis, the intake faced into a semi-
curacy. The CFD domain included the engine and external flow sphere. Due to a region of very sheared cells just upstream of the
and was split into two main blocks: The upstream block included intake, a rectangular domain proved beneficial. The size of the
the intake and external flow of the forebody and the downstream domain was investigated in the direction upstream of the engine
block included the bypass duct and nozzle and the external flow of and perpendicular to the axis, selecting the height 共i.e., radius兲 of
the afterbody. These blocks could be detached and calculated in- the axisymmetric domain.
dividually, but for the present study, the full domain was consid- The domain length upstream of the engine face had a relatively
ered in all cases. small influence on the pressure recovery of the intake. Similar
The commercial flow solver “FLUENT” and its preprocessor levels were predicted for domain lengths of eight times the fan
“GAMBIT” were used to provide the CFD modules 关17兴. GAMBIT diameter D2,tip for different grid densities. The node density up-
allowed for generating structured grids and featured boundary stream was investigated for a wider range and a plateau of the
layer functions for easy application at walls. It was optional to pressure recovery factor was reached for densities in excess of 12
map the boundary layer region and the main grid independently to nodes/m.
the computational domain boundaries, which facilitated the gen- The height of the domain was investigated in a similar manner.
eration of a flexible multiblock structure within an individual However, it was an aim that, once a CFD module was established,
mesh block. it could be applied to any operating point. The nature of the flow
Figure 5 shows two versions of a 2D mesh block of a cross at the engine intake varies significantly between the extreme con-
section of an aeroengine intake duct. In 共a兲, a thicker boundary dition of static operation and top speed. The essential change is
layer was created compared to 共b兲. To provide adequate resolution the mass flow ratio 共MFR兲 and has been well described by Seddon
of the boundary layer, the boundary layer meshes were attached to and Goldsmith 关12兴:
the solid boundaries of the computational block. This body-fitted
• At ground running 共MFR→ ⬁ 兲 the streamlines into the fan
approach increased numerical accuracy as the flow was well
aligned with the mesh. It resulted in a C structure around the face may originate from upstream or downstream of the in-
take. In computational terms the flow might enter from the
front, top, or rear of the domain. This raises issues for the
definition of boundary types. Furthermore, the flow may not
be able to follow the strong curvature around the intake lip
and flow-separation zones may appear. This results in a sig-
nificant reduction of the pressure recovery factor of the in-
take.
• At lower flight speeds during climb a higher power setting
may lead to an operation in suction mode 共MFR⬎ 1兲.
• For high speed cruise the intake is generally designed such
that the internal duct performance is at its optimum. The
Fig. 5 Structured mesh block using the boundary layer func- intake operates in spillage mode in a typical range of
tion to create multiblocks MFR= 0.5. . . 0.8. At such low mass flow ratios separation is

764 / Vol. 129, JULY 2007 Transactions of the ASME

 Fig. 7 Pressure recovery factor of the bypass-section versus
Fig. 6 Pressure recovery factor for different domain heights at radial nodes

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

static conditions

turbulent viscosity were solved using the second order upwind

scheme.
avoided and the pressure recovery factor at its maximum
within the flight envelope. All streamlines entering the fan CFD Module of the Bypass. Similar considerations had to be
face originate from the upstream boundary of the computa- taken into account to set up the numerical model for the bypass
tional domain. and nozzle. Using a rectangular domain for the study, the neces-
• Under top speed conditions, the mass flow ratio can reduce sary length of the domain downstream of the engine needed to be
even further. established. As only the pressure loss within the low-pressure sys-
tem was considered in this study, the dependency of this loss was
Generating a computational domain coping with the entire investigated for different domain lengths. For the purpose of this
flight envelope, required that the domain height in particular was study a domain length of around 16 Dmax was considered to be
examined for static conditions. The pressure recovery factor ver- sufficient, which enabled the prediction of the bypass pressure
sus the domain height for sea-level-static conditions is shown in loss to the fourth digit.
Fig. 6. The height is referred to the maximum nacelle diameter It was essential to provide a mesh within the bypass duct that
Dmax. For heights above 8Dmax the pressure recovery factor was produced grid-independent predictions of the pressure loss. There-
predicted to the fourth digit. A plateau appeared for node densities fore, the overall pressure loss from fan exit to the bypass-nozzle
in excess of 20 nodes/m. However, the pressure recovery was exit was calculated for different numbers of radial nodes along the
predicted to the fourth digit with a deviation of 0.003% relative to bypass height. The results are shown in Fig. 7. The slope indicated
the case of higher node density. This was considered to be close that for 60 radial nodes, the pressure recovery for the low-pressure
enough to the plateau values and a node density of 15 nodes/m system was predicted to the third digit, possibly the fourth. A
was found to be sufficient for further cases. further increase of radial nodes would have required a further grid
The necessary domain height for conditions at altitude was sig- refinement in the streamwise direction to avoid high aspect ratios
nificantly smaller with in the region of 4–5.5Dmax in spillage for the cells near the wall. This would have increased the mesh
mode. In order to save computational time for the investigated size significantly. A mesh with 60 radial nodes was therefore cho-
cases under cruise and end of climb conditions, using a domain sen for the study.
height of 5.5Dmax reduced the grid size by 26%. The boundary It has to be stated that the numerical model of the bypass did
conditions were set as pressure conditions and were derived from not include the outlet guide vanes of the fan and swirl was not
the predicted values of the performance simulation. For the top of included in the model. The predicted loss was therefore small. For
the domain, however, a symmetry condition which aligned the future investigations, a porous pressure jump could be included at
flow with the boundary proved favorable for engines at higher the location of the outlet guide vanes to model a pressure reduc-
speeds. For the static conditions the larger domain was used and tion at the bypass entry.
the boundary type at the top of the domain was changed to a
pressure inlet condition in order to allow streamlines from this
boundary to enter the engine inlet. Matching of Performance and CFD Simulations
The density of the grid in the internal duct of the inlet was The idea of combining CFD modules and performance simula-
investigated for a range of 60–90 nodes along the span of the fan. tion required an iteration procedure. The CFD module supplied
A plateau of predicted pressure recovery factor was indicated for with ideal boundary condition, predicted the pressure loss for a
90 or more nodes along the span. As a higher number of nodes led certain mass flow. The performance code simulated the mass flow
to higher aspect ratios of the cells far upstream, 90 nodes were reduction due to the pressure loss after the engine rematched it-
considered to be sufficient. A prediction of the pressure recovery self. Therefore, the pressure loss and mass flow had to be matched
factor to the third digit was reached for 70 and more nodes. The between the two tools and this was achieved by manual iteration.
prediction of the fourth digit was expected with 90 nodes along As the performance tool provided only a total flow condition
the span. and the CFD modules required static conditions at the pressure
Turbulence was modelled for all the CFD modules using the exits, the static conditions were calculated using the Q function.
Spalart–Allmaras model, which is implemented in FLUENT using For the first CFD simulation, the boundary conditions from the
wall functions. The turbulence intensity was set to 1%. This com- performance tool were taken directly. For further iterations, a flow
bination resulted in realistic profiles at the fan inlet, whereas discharge coefficient based on the last CFD iteration was included
higher turbulence intensities showed local pressure overshoots at to estimate the required static pressure. Matching for the intake
the boundary layer edge. Other turbulence models showed ex- was usually achieved within ten iterations and a typical conver-
ceedingly higher losses or comparable results at the price of gence plot of the matching procedure is shown in Fig. 8.
higher computational effort. The residuals were calculated based on the deviation from the
All cases have been performed with the coupled implicit for- previous iteration step. The even numbers of iteration steps rep-
mulation of FLUENT. The discretisation schemes for the flow and resented results from a CFD simulation, whereas the odd numbers

Journal of Engineering for Gas Turbines and Power JULY 2007, Vol. 129 / 765
 Fig. 9 Key whole-engine performance data for conventional
0D and 2D performance simulation using CFD modules

Fig. 8 Residuals for intake matching between performance For cruise conditions, the intake loss is at a relatively low level.

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

and CFD tools The resulting effect on overall engine performance was expected
to be equivalently small. However, the results show that the cost-
effect ratio is greater than one in terms of thrust degradation.
Along with an increased SFC, the thermodynamic effect is there-
stood for performance predictions. It can be seen that the pressure fore worth considering.
residuals reach zero with every performance prediction, but the The radial fan-inlet profile of the matched solution is shown in
pressure loss had to be recalculated using the CFD module. When Fig. 10 共left兲. The undistorted fan pressure ratio profile and the
both residuals were close to zero, a matching was achieved. For resulting profile when the CFD intake module and the engine
the present study, the convergence criteria was set to 0.01% de- performance model were matched are illustrated on the right hand
viation in mass flow and 0.0% in total pressure. side. As might be expected intuitively, the tip distortion had the
The fan-exit profiles for the total pressure and total temperature largest effect on the redistribution of the pressure ratio.
were provided from the performance tool. This implied that for The prediction of hub distortion caused by the spinner had a
the bypass duct, the level of static pressure at the core inlet con- smaller effect on the pressure ratio of the fan. It should be stated,
trolled the fan-exit mass flow and the core-inlet mass flow. The that the inclusion of the boundary layer resulting from the spinner
bypass nozzle was running at close to choked conditions and may be questionable: In case that the radial profiles for the fan
therefore the changes of bypass mass flow were small. The match- performance component originated from test data, these would
ing procedure was more complex and required slightly more, with include this effect already. The performance of a fan cannot be
up to 15 iterations 共equivalent to seven CFD simulations兲. measured without a spinner present and the consideration of the
The fan-exit profiles provided from the performance code were intake boundary layer caused by the spinner would lead to a
defined for 19 radial stations, whereas the grid-independent CFD double booking of these losses. As notional data were used for the
module counted 90 nodes along the fan-exit span. To provide present case, hub, and tip distortion were considered. A simple
approximate data for each node, curve fitting was performed and solution to overcome the double booking when a CFD module for
P3共r兲 and T3共r兲 was derived for each CFD node. This was initially the intake was used, was to assume unity for the pressure recovery
done in EXCEL spreadsheets, but later integrated into a MATLAB factor 共PRF兲 in the hub boundary layer or to continue the peak
procedure to ease user application. pressure of the profile to the hub.
To take advantage of the losses provided by the CFD simula-
tions for the bypass and bypass nozzle, the 0D bypass-nozzle Discussion
component of the performance code was changed such that the
discharge coefficient and pressure loss predicted from the CFD This study demonstrated the feasibility of assisting gas turbine
was used instead of the original characteristic. The losses of the performance simulation with CFD methods to achieve a more
bypass duct and bypass nozzle were combined, which required to realistic engine model. The initial effort that had to be invested to
derive the mass-averaged total pressure at the bypass inlet only. design the intake and bypass components, including the numerical
For the present study, this was evaluated using an EXCEL spread- investigations to assure that the CFD modules were numerically
sheet for the fan-exit total pressure distribution and the radial appropriate, was relatively high. However, it allowed the engine
mass flow distribution 共from the CFD simulation兲. Based on the performance to be investigated with a more realistic, higher order
bypass ratio predictions of the performance model, the coordinate of dimension and was therefore considered to be applicable over a
of the split between the bypass and core flow was calculated and wider operating range than conventional characteristics-based
the mass-averaged total pressure for the bypass section was de- methods. The design procedure can be easily adapted for other
rived. high-bypass turbofans and for similar bypass ratios, a scaling of
A perfect matching between the CFD modules and the perfor- the generated CFD modules may be possible. The structure of the
mance code for the bypass duct was not achieved due to mass
flow discrepancies between the radial fan performance model and
the CFD simulations. The best match for cruise conditions showed
mass flow mismatches of 0.54%, 0.15%, and 0.99% for the bypass
duct, core inlet, and fan exit, respectively.

Performance Results
The engine performance results for cruise conditions are shown
in Fig. 9. Including the CFD predictions of the inlet pressure loss
of 0.44% into the model, resulted in a reduction of thrust by
1.02% and an increased SFC of 0.29%. The additional 2D mod-
eling of the bypass loss resulted in a pressure loss of 0.43% which
reduced the thrust by a further 0.49% and resulted in a further Fig. 10 Radial profiles of the intake pressure recovery and un-
increase of SFC of 0.10%. distorted and distorted fan pressure ratio

766 / Vol. 129, JULY 2007 Transactions of the ASME

modules allows for changes of the boundary layer mesh which show a greater than one cost-effect ratio in terms of thrust.
should enable an adaptation to different flow conditions. Therefore, it appears well worth considering.
The CFD modules were a first step to a higher-fidelity model of • The computational effort that had to be invested to achieve
the intake and bypass system, but were still based on major sim- the results was quite high. A matched result between the
plifications which limited the application to certain scenarios. Un- performance code and the CFD modules required several
der the assumption of axisymmetry, only inlet flow parallel to the days. This is most likely to be outside the time scopes in an
engine axis can be considered, allowing for radial distortion mod- industrial environment. However, previous experiences can
eling. The more severe types of circumferential distortion, for be used to improve initial guesses and may speed up the
example in crosswinds, cannot be modelled using this 2D ap- procedure such that it could be performed within one day.
proach.
In terms of the bypass, the CFD modules could not include the
outlet guide vanes or the pylon. The predicted losses were there-
fore only due to friction losses in the “empty” duct and were Acknowledgment
significantly lower than in real engines. Also, swirl, which would Thanks go to Darrell D. Williams for his kind advice regarding
normally enter the bypass from the fan, though compensated to a the design of the intake.

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

degree by the outlet guide vanes, has not been taken into account.
A representative to the loss generated in the outlet guide vanes
could be included into the model by using a so-called “porous
pressure jump.” This, in effect, calculates a pressure decrease in-
troduced by a membrane of zero thickness as a function of per- Nomenclature
meability, thickness, and a pressure jump coefficient. In the ab- a ⫽ stream-wise distance from highlight to throat
sence of validation data, this appeared to be a crude approach and b ⫽ rH − rth
was therefore not considered in this study. BLE ⫽ boundary layer edge
Even though this 2D approach for the CFD modules may ap- BPR ⫽ by pass ratio
pear very simple, the mesh for the intake and bypass system CFD ⫽ computational fluid dynamics
yielded in excess of 100 k cells. On a single processor, the first CR ⫽ contraction ratio
CFD calculation of an operating point usually required around D ⫽ diameter
12 h. The boundary conditions of the far field had to be adjusted DR ⫽ diffusion ratio
in several steps to the final values to avoid immediate divergence. MFR ⫽ mass flow ratio
This required user attention during the first 1–2 h. The first con- NPSS ⫽ numerical propulsion system simulation
verged CFD solution was then taken as the initial solution for the P ⫽ total pressure
next iteration between the performance code and the CFD simu- PR ⫽ pressure ratio
lations. These cases converged faster due to the closer proximity PRF ⫽ pressure recovery factor
of the initial to the final solution. The next solution was achieved r ⫽ radius
generally within 4–6 h. With a growing database of operating SFC ⫽ specific fuel consumption
points for an investigated engine model, the convergence time of T ⫽ total temperature
the CFD modules could be reduced and under best conditions, the TET ⫽ turbine entry temperature
time to reach a solution for a certain operating point may well be ␤ ⫽ boat-tail angle
within one day. Subscripts
1 . . . 10 ⫽ station numbers
a ⫽ afterbody
Conclusions
f ⫽ forebody
• To overcome the lack of geometrical data, fundamental de- gr ⫽ gross
sign principles were applied to create a generic, but repre- H ⫽ highlight
sentative intake and bypass geometry. The results for each LPS ⫽ low-pressure system
component appeared reasonable and the design process can th ⫽ throat
be adapted easily to other engines. The general purpose of
using such data on this occasion was not related to achiev-
ing specific design improvements. A data set was required as
a vehicle to demonstrate performance trends. In the aca-
References
关1兴 Hall, E. J., Delaney, R. A., Lynn, S. R., and Veres, J. P., 1998, “Energy
demic environment, the lack of detailed design information Efficient Engine Low Pressure Subsystem Aerodynamic Analysis,” Technical
is often an issue. But in similar cases, this simplified design Memorandum TM-1998-208402, NASA.
procedure may help to overcome this hurdle. 关2兴 Riegler, C., Bauer, M., and Kurzke, J., 2001, “Some Aspects of Modeling
• The CFD modules for the intake and bypass section were Compressor Behavior in Gas Turbine Performance Calculations,” Very Large
Data Bases, 123共2兲, pp. 372–378.
generated with considerable effort due to a software bug in 关3兴 Curnock, B., Yin, J., Hales, R., and Pilidis, P., 2001, “High-Bypass Turbofan
the preprocessor of the CFD software. A meshing approach Model Using a Fan Radial-Profile Performance Map,” Aircraft Design, 4共2–
of structured grids with boundary layers used to provide 3兲, pp. 115–126.
multiblock structures can not be advised using GAMBIT. Al- 关4兴 Marshall, C. D., 1998, “The Modelling of High Bypass Turbofans,” Master’s
thesis, Cranfield University, UK.
ternatively, a different mesh generator or an unstructured 关5兴 Yin, J., and Pilidis, P., 2002, “Influence of Inlet Profile on High-BPR Turbofan
approach may be used when starting a CFD simulation from Performance Using a Radial Profile Map,” Proceedings of the 23rd Interna-
scratch. tional Council of Aerospace Sciences, Toronto, Canada, Sept. 6–13 2002,
• The combination of performance simulation with CFD mod- ICAS 2002-R79.
关6兴 Li, M. S., 2004, “2D Low Bypass-Ratio Turbofan Modelling,” Ph.D. thesis,
ules for the intake and bypass system has shown correct Cranfield University, UK.
qualitative results. In the absence of information about pres- 关7兴 Cumpsty, N., 2000, Jet Propulsion, Cambridge University Press, Cambridge,
sure losses for these components, it is expected that valid UK.
results can be achieved for a wider operating range using the 关8兴 RTO, 2002, “Performance Prediction and Simulation of Gas Turbine Engine
Operation,” RTO Technical Report 044, NATO Research and Technology Or-
CFD modules. ganisation, Neuilly-Sur-Seine Cedex, France.
• Radial inlet distortion affects the whole-engine perfor- 关9兴 McKenzie, A. B., 1997, Axial Flow Fans and Compressors, Cranfield Series
mance. The thermodynamic effects may seem small, but on Turbomachinery Technology, Ashgate Publishing, Aldershot, UK.

Journal of Engineering for Gas Turbines and Power JULY 2007, Vol. 129 / 767
关10兴 Albers, J. A., and Miller, B. A., 1973. Effect of Subsonic Inlet Lip Geometry plant research report IAR 85002, Rolls Royce, UK.
on Predicted Surface and Flow Mach Number Distributions, Technical note 关14兴 ESDU, 1994, NACA 1-Series Geometry Representation for Computational
NASA TN D-7446, NASA. Fluid Dynamics, Data item 94013, ESDU International, London, UK.
关11兴 ESDU, 1981, Drag of Axisymmetric Cowls at Zero Incidence for Subsonic 关15兴 Williams, D., 2004, “Propulsion Systems Performance and Integration,”
Mach Numbers, Data item 81024, ESDU International, London, UK. “Thermal Power,” Cranfield University, Lecture Series Cranfield, UK.
关12兴 Seddon, J., and Goldsmith, E. L., 1999, Intake Aerodynamics, 2nd. ed., Black- 关16兴 Walsh, P. P., and Fletcher, P., 1999, Gas Turbine Performance, Blackwell
well Science, Oxford, UK. Science, Oxford, UK.
关13兴 Stanhope, F. W., 1968, “The Performance of NACA 1-Series Intakes,” Power 关17兴 FLUENT, 2003, FLUENT 6.0 Documentation, FLUENT Inc.

Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/129/3/761/5650678/761_1.pdf by University of Sheffield user on 07 June 2024

768 / Vol. 129, JULY 2007 Transactions of the ASME