Advances in Engineering Software 68 (2014) 9–18

Contents lists available at ScienceDirect

Advances in Engineering Software

journal homepage: www.elsevier.com/locate/advengsoft

Integrated aerodynamic design and analysis of turbine blades

Chengen Wang ⇑
Northeastern University, Shenyang 110819, China

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents an integrated approach for aerodynamic blade design in an MDO (multidisciplinary
Received 16 August 2013 design optimization) environment. First, requisite software packages and data sources for flow computa-
Received in revised form 19 September 2013 tions and airfoil modeling are integrated into a single cybernetic environment, which significantly
Accepted 27 October 2013
enhances their interoperability. Subsequently, the aerodynamic blade design is implemented in a
Available online 21 November 2013
quasi-3D way, supported by sophisticated means of project management, task decomposition and allot-
ment, process definition and coordination. Major tasks of aerodynamic blade design include 1D meanline
Keywords:
analysis, streamsurface computations, generation of 2D sections, approximation of 3D airfoils, and 3D
Aerodynamic blade design
Airfoil modeling
flow analysis. After compendiously depicting all the major design/analysis tasks, this paper emphatically
Design integration addresses techniques for blade geometric modeling and flow analysis in more detail, with exemplar
Design collaboration application illustrations.
Engineering process management Ó 2013 Elsevier Ltd. All rights reserved.
Flow analysis

1. Introduction limits. While, mechanics engineers are expected to precisely pre-

Turbine blades are core components of an aeroengine, a typical
sub-category of gas turbines, which transform aerodynamic and
thermal energy carried by gas fluids into mechanical driving en-
ergy. Consequently, operations of turbine blades largely dominate
the global engine performance, such as thrust, aerodynamic and
thermal efficiency, fuel-efficiency, and reliability. Today’s increas-
ingly augmented requirements for high performance aeroengines
subject turbine blades to fiercely intensive aerodynamic, thermal
and mechanical loadings. These extreme physical loadings make
turbine blades work in very harsh conditions that approach the
limits of its material properties such as melting points, stress and
damage-resistance thresholds.
Accordingly, turbine blade design has been regarded as one of
the most difficult engineering problems, which has drawn inten-
sive attention from both industrial and academic circles. Generally,
design of turbine blades is a complex multidisciplinary process
involving the integration of several disciplines such as aerodynam-
ics, structures, dynamics, and heat transfer [1].
Disciplinarily, aerodynamics engineers have to control the very
complex flow phenomena occurring in highly loaded stages, on the
whole operating range of the engine [2]. And, thermodynamics
engineers have to accurately comprehend the heat transfer pro-
cesses between gas flows, coolants and solid structures, and keep
temperatures within the blade structure well below the allowable
⇑ Tel.: +86 24 83683619.
E-mail address: wangc@mail.neu.edu.cn
dict and simulate structure deformations and dynamic responses
under the aerodynamic, thermal and mechanical loadings.
Conventionally, the multiple disciplinary computations were
carried out separately by different groups of engineers, aided by
a number of stand-alone problem-solvers. Conceivably, the iso-
lated computing processes result in remarkable engineering ineffi-
ciency and ensnarement in local optimums. Besides, the isolated
legacy problem-solvers leave out of account strong couplings
among disciplinary computations for turbine blade engineering.
To effectively tackle the drawbacks of traditional sequential de-
sign approaches, companies are growingly employing integrated,
collaborative, and multidisciplinary design technologies. Actually,
since the advent of information technologies, researchers have
been continuously looking for cost-effective integration ap-
proaches to build bridges among isolated ‘‘automation islands’’
[3,4].
In regard to aeroengine design, a variety of integrated engineer-
ing infrastructures have been implemented to improve interopera-
bility among heterogeneous software packages and data sources.
For instance, Talya et al. (2000) presented an integrated multidis-
ciplinary design optimization procedure for design of both inter-
nally and externally cooled gas turbine blades [5]. Houstis et al.
developed a multidisciplinary problem solving environment to
support collaborative design of gas turbines, where the legacy FOR-
TRAN and C codes were wrapped up and delegated by agents [6].
As a common practice, turbine blade design commences with
aerodynamic shape design, of which the ultimate goal is to find
the airfoil surfaces that optimize the blade’s aerodynamic

0965-9978/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.advengsoft.2013.10.005
10 C. Wang / Advances in Engineering Software 68 (2014) 9–18
performance. Blade airfoil design essentially relies on numerical
computations that describe and predicate various fluid flows with-
in turbine stages, in compliance with governing equations of mass,
momentum and energy conservations. The fluid governing equa-
tions are generally called Navier–Stokes equations in the literature,
and the corresponding software packages are called 3D Navier–
Stokes solvers.
Thanks to rapid enhancements of computational capabilities,
software codes of finite difference methods (FDM), finite element
methods (FEM) and finite volume methods (FVM) are able to solve
the 3D governing equations in reasonable time periods. However,
finding numerical solutions to the governing equations in 3D blade
passages is so time-consuming that for a long time the equations
have been solved in 2D or quasi 3D spaces by industries.
And, in engineering practice, many companies have not shifted
to the full 3D computational paradigm; 2D and quasi-3D aerody-
namics solvers still play a dominant role in blade engineering. Re-
cently, Koini et al. (2009) developed a software tool for
interactively constructing parametric 3D blade models on basis
of 2D profile sections [7]. Qiu et al. (2010) presented an integrated
blade design system that incorporates 1D meanline analysis, quasi-
3D throughflow and blade-to-blade flow calculations, and 3D CFD
[8].
In accompaniment to worldwide zeal for multidisciplinary de-
sign optimization (MDO), we have been developing an MDO envi-
ronment [9] for a decade, which concordantly incorporate
information and application integration techniques [10–13], opti-
mization algorithms [14], response surface models [15], etc. The
MDO environment is a platform of high maintainability, scalability
and flexibility, which may accommodate existing legacy problem-
solvers and emerging problem-solvers. With support from a num-
ber of manufacturing companies, the ad hoc MDO environment has
been employed in several engineering analysis and design scenar-
ios like turbine blade design, aircraft flight-load design, aeroengine
heat transfer analysis, ship power system design, and develop-
ments of car molds.
The goal of this paper is to present how the MDO environment
is implemented to aerodynamic turbine blade design, by perform-
ing ‘‘what-if’’ computations of flow velocity and pressure distribu-
tions, and blade geometries. Specifically, this work addresses an
initial pilot phase of implementing MDO techniques to aeroengine
design, which aimed at validating feasibility and merits of applying
multidisciplinary design integration and optimization techniques.
To reduce implementation complexity and to rapidly gain tech-
nical benefits, the practitioner engineers decided that emphasis of
the pilot phase was put on design integration and collaboration in
the context of aerodynamic blade engineering. This meant that the
pursuit of design optimization was not explicitly addresses in this
stage, and perspectives of heat transfer analysis and structural
analysis were not accounted.
The rest of this paper is structured as follows. Section 2 de-
scribes how software codes and data sources used for quasi-3D
aerodynamic blade design are integrated. Section 3 first explains
how aerodynamic design is defined and managed in terms of pro-
jects, tasks and workflows. Then, Section 3 depicts the major tasks
of blade airfoil engineering. Section 4 depicts in more detail the
geometric modeling techniques that are used to approximate blade
airfoils. Section 5 presents how flow characteristics analysis, vari-
ous losses and efficiencies are computed in the MDO environ-
ments. Section 6 finally concludes this paper.
2. Aerodynamic blade design integration
In practice, a turbofan engine may run with one, two or up to
three spools, which respectively have one, two or three turbines
for extracting energy from the exhaust fluid. In turn, a turbine
comprises multiple stages, each of which has a rotor that extracts
the fluid energy, and a stator that adjusts the flow velocity and
direction. Compositionally, a rotor is a cascade of blades attached
to a disc at the roots, in which the throughflows of exhausts mainly
move forwards in parallel to the axis of rotation. And, a stator is a
cascade of blades or vanes attached to the engine casing at the tips,
in which the throughflow velocity and direction are adjusted. This
work addresses integrated aerodynamic blade design through all
turbine stages in the MDO environment.
As unable being computed by a monolithic software package,
the quasi-3D blade aerodynamic design is collectively computed
by over a dozen of disciplinary analysis and design codes that
respectively predict 1D, 2D and 3D flow characteristics, and depict
the corresponding blade geometries in 2D and 3D spaces. It is
implicitly assumed that the aerodynamic computations are con-
ducted without accounting for heat transfer (adiabatic conditions),
and structural responses induced by the aerodynamic loads.
2.1. Integration of blade aerodynamic design codes
Noticeably, all the in-house developed 1D and 2D flow comput-
ing codes and blade profile modeling codes were written in proce-
dural FORTRAN and C languages, which use plain text files for data
retrieval and transfer. And, engineers had to manually prepare the
input files and visually parse the output files of computation pro-
cesses because the legacy software applications were default of
Graphical User Interfaces (GUIs). Generally, these stand-alone
problem-solvers did not support out-process communications
and multiple concurrent user access.
Consequentially, integration techniques are implemented as an
essential part of the MDO environment, which accommodate all
the codes and data sources on a single cybernetic platform. This
enables effective coordination and tight collaboration in design of
turbine blade airfoils.
Application integration allows executions of disciplinary soft-
ware packages to be effectively controlled and coordinated. In
the MDO environment, aerodynamic engineers do not have to
memorize a large number of command lines for launching the leg-
acy problem-solvers. And, they are provided with means to inter-
act with the computation processes, by using surrogate objects
[10].
Software packages integrated into the MDO environment are no
longer independent systems but parts of a whole. Therefore, the
software packages need to be known of their existence and let their
executions be controlled and coordinated. The MDO environment
uses dynamic registration and configuration approaches to keep
track of the integrated applications. As shown in Fig. 1, an applica-
tion’s registration information generally includes its name, disci-
pline, command lines, versions, running environments,
executable file with a suffix of exe or bat, numbers and names of
the input/output files, its host machine and absolute path.
2.2. Interoperation of heterogeneous data sources
In conjunction with application integration, information inte-
gration carries out interoperations among heterogeneous data
sources, by building mapping relations among data objects, seman-
tically parsing unstructured input/output files, and transforming
data representation schemas. For instances, data mapping methods
are implemented to discern disciplinary parameters contained in a
solver’s input/output files, and to maintain dynamic linkages be-
tween the non-structured parameter files handled by a solver
and the structured data objects (database records or XML files)
handle by the integration platform [11].
 C. Wang / Advances in Engineering Software 68 (2014) 9–18
Fig. 1. Integration registration of a software package.
Information integration relieves engineers from frequently
switching computation processes and transferring data forth and
back among them. Information integration provides engineers
with appropriate means to keep records of design evolutions and
to analyze the cause-effects between input and output parameters.
Information integration also enables capturing and reusing engi-
neering knowledge and know-hows accumulated by specialist
engineers in long time courses.
Technically, information integration implementations are ser-
vice-oriented, which implies that all operations and transforma-
tions performed on data objects are regarded as services
provided by mediators. Similarly, Yu et al. presented a CAE-
integrated distributed collaborative design system that encapsu-
lated CAE software tools by service providers [16]. In this work,
the ActiveX Data Objects (ADO) technology is used for handling
both relational and non-relational data sources. ADO components
abstract the semantics of the data transactions and provide inter-
faces for retrieving and performing operations on the stored data.
In the MDO environment, information integration techniques
provide communications between designers and applications,
and also automate data transfers between one application and an-
other. For instance, Windows based user interfaces hide an appli-
cation’s command lines and permit engineers to input and select
design parameters via well-structured tables, views and pop-up
dialogs.
In summary, application and information integration addresses
human factors, applications and data sources in a holistic way,
which considerably enhances interoperability among distributed
engineers, dissimilar software packages, and heterogeneous data
sources.
To avoid descriptive redundancy, we refer readers to our previ-
ous publications for technical details of the integration techniques
[3,9–11].
11
3. Aerodynamic blade design workflow management
Integration sets requisite basis for aerodynamic blade design to
be smoothly streamlined into workflows and thus coordinated in
alignment with well-established procedures. Based on integration
of disciplinary codes and data, the MDO environment provides
sophisticate project and workflow management techniques to sup-
port design integration and collaboration.
As shown in Fig. 2, the MDO environment provides means to
decompose blade aerodynamics design and analysis into a number
of tasks, which are deliberately defined in terms of actors, prece-
dence restraints and resources. Subsequently, workflows of aero-
dynamic blade design are built on basis of tasks, their pre- and
post-conditions, triggering events, etc. Undoubtedly, workflow
management techniques are of considerable significance to
enhancing blade design efficiency and integrity.
In a general sense, aerodynamic blade design is a progressive
process evolving from one-dimensional (1D) design, through 2D
profile curvature modeling, to 3D surface approximation and flow
performance analysis. More specifically, aerodynamic blade design
and analysis is composed of computations of 1D meanline analysis,
2D streamsurfaces, basis section profiles, stacking, 3D airfoil mod-
eling, 3D flow simulations, and prediction of efficiencies/losses.
Conceivably, a generic blade aerodynamic design process is an
aggregation of several sub-processes, which are composed of a
number of tasks. Logically, each design/analysis sub-process is a
recursive loop of computational tasks, and, the global blade design
process is an iterative loop of the sub-processes. Below, we com-
pendiously discuss the major sub-processes of aerodynamic blade
design and analysis.
Due to length limit, this section will not further address the
workflow issues because detailed workflow specification, verifica-
tion, and control techniques can be found in previous literature,
12 C. Wang / Advances in Engineering Software 68 (2014) 9–18
Fig. 2. Aerodynamic blade design task management.
such as the requirement-oriented automated framework for formal
verification of service workflows [17]. Quite recently, Lwin et al.
also developed a Web-based MDO framework that aimed at simul-
taneously integrating legacy code and workflow management,
which were used for helicopter rotor blade design [18].
3.1. One-dimensional meanline analysis
Despite the sophistication and improvement of 3D CFD technol-
ogy, one-dimensional meanline analysis still has its irreplaceable
role in the turbomachinery design process [8]. Consequentially,
aerodynamic blade design often begins with one-dimensional or
1D meanline design/analysis that calculates global flow parame-
ters along mid-height line of blades through the turbine stages.
1D meanline analysis calculates aerodynamic performance of
the exhaust flow along a single streamline at mid-heights of blades
through all turbine stages. Leaving vortex or rotational flows out of
account, 1D meanline analysis depicts the velocity of the exhaust
flow in two velocity vectors, absolute and relative velocities.
Assuming constant axial flow velocity and meanline blade radius,
the meanline analysis calculates velocity triangles at blade leading
edges and trailing edges [19].
The objective of 1D meanline analysis is to determine blade/
vane inlet and exit angles, pressure ratio per stage, the number
of stages, numbers of blades and vanes of each stage, and annulus
blade shapes, which provides inputs to subsequent calculations.
Normally, the exhaust flow from a combustor first enters a sta-
tor, called an inlet guided vane (IGV), the utility of which is to
ready the flow velocity and direction. Then, the exhaust flow exit-
ing the IGV with an absolute velocity enters the first stage rotor,
which rotates with an angular velocity. The absolute velocity of ex-
haust fluids, the circumferential velocity at the mid-height of a
blade, and the relative velocity of the fluids to blade constitute
the well-known velocity triangle.
Through the first stage rotor the fluids transfer energy to the
blades, which decreases fluid pressure and density. While the axial
component of flow velocity remains constant, the circumferential
component of flow velocity increases. After leaving the rotor, the
fluids enter the first stage stator that regulates the flow velocity
and direction for the second stage rotor. And, similar changes in
fluid velocity, pressure and density take place in the subsequent
turbine stages.
The velocity triangles enable engineers to select desirable inlet
and exit angles at meanline radii of blades and vanes. 1D meanline
analysis also helps engineers to make decisions on several critical
aerodynamic parameters. For instance, fluid pressure drops
through a rotor because of fluid expansion and energy extraction.
Conceivably, pressure ratio of inlet pressure to outlet pressure in
a rotor blade is of great significance to the aerodynamic efficiency
and turbine structure. Pressure ratio in a turbine cascade is gener-
ally set larger than that in a compressor cascade, which permits
less turbine stages to extract sufficient energy.
One dimensional meanline analysis helps engineers to choose
pertinent values for a number of aerodynamic parameters. Cate-
gorically, flow coefficient determines the ratio of the axial velocity
of fluids to circumferential velocity at a blade mean radius, which
geometrically dominate flow turning angles. This parameter exerts
impacts on mass flow rate and hence the blade throughflow areas.
Stage loading coefficient is another decision parameter that deter-
mines the number of stages in a turbine, in conjunction with mass
flow rate and required energy output.
In regard to the number of blades in a cascade, the pitch-to-
chord ratio or solidity is a crucial parameter.
As axial velocity and blade radius remain constant throughout
turbine stages, whereas gas density decreases down turbine stages.
To counteract the effect of reduction in fluid density, the blade
throughflow area, the annulus area, has to progressively increase
from stage to stage. This leads to considerable increase in blade
heights through stages.
 C. Wang / Advances in Engineering Software 68 (2014) 9–18 13

As shown in Fig. 3, the 1D meanline analysis yields preliminary

design of turbine meridional flow path, number of stages, flow
characteristics, etc. It is observed that there exist remarkable dif-
ferences between blade numbers and shapes among turbine
stages.

3.2. Streamsurface computations

After performing 1D meanline analysis, engineers distribute

geometric parameters along the radii, and proceed to perform
two dimensional or 2D streamsurface analyses.
Two-dimensional aerodynamic blade design uses some stream-
surface algorithms to ascertain the blade cross-section flows and
throughflows that fulfill the aerodynamic requirements. In 2D
aerodynamic computations, streamsurfaces, especially the so-
called S1/S2 streamsurfaces, are formulated to control velocities
and pressures of flows and determine blade profiles in terms of Fig. 4. Illustration of S1/S2 streamsurfaces.
coordinates of critical points and surface curvatures.
The definition of blade-to-blade (S1) and hub-to-tip (S2) stream
The S1 streamsurfaces, generally assumed surfaces of revolu-
faces was introduced by Wu and this viewpoint dominated the
tion, calculate flows from a blade to its adjacent blade. Like the
subject until the early 1980s when fully three dimensional (3D)
S2 streamsurfaces, S1 streamsurfaces are also solved in two modes.
methods first became available [20].
S1 solvers either determine blade section profiles with given sur-
Fig. 4 presents an illustration of the S1/S2 streamsurfaces. The
face pressure distributions (the inverse design mode), or calculate
axisymmetric S2 streamsurfaces calculate hub-to-tip through-
blade surface pressure distribution with given blade section
flows, which is essential to work outputs of turbomachines. The
geometries.
natural formulation of the throughflow problem is to obtain the
There are many available numerical methods, such as stream-
solution with a specified mass flowrate [20]. During the design
line curvature and stream function methods, are applied to solve
process, the S2 streamsurfaces are solved in direct and inverse for-
the S1/S2 stream surfaces. In this work, in-house developed ad
mulations. The direct S2 streamsurface solver calculates the flow
hoc S1/S2 computation codes are applied to solve the direct and in-
velocity distribution when blade geometric variations are given.
verse streamsurface problems.
The inverse S2 streamsurface solver determines blade spanwise
variations with given velocity distributions.

Fig. 3. Turbine meridional path.

14 C. Wang / Advances in Engineering Software 68 (2014) 9–18
3.3. Blade airfoil design
The major task following up 1D and 2D flow analysis is to de-
sign blade airfoils, in terms of 2D cross section profiles and 3D solid
geometric models. As blades are dissimilar through stages, blade
airfoil computations are repeatedly performed for each turbine
stage.
With respect to designing a single baled airfoil, several cross
section profiles, often known as basis sections, are first developed
in accordance with calculations S1/S2 streamsurfaces. Design of a
blade section profile attempts to attain curvature smoothness
and reduction of profile losses. The profile or airfoil section is com-
posed of four distinct curves, corresponding to the suction side, the
pressure side, leading and trailing edges, and must be curvature
continuous, at least retaining C2 continuity [21].
In this work, a number of interpolation and approximation algo-
rithms are implemented to determine appropriate blade section
profiles. As shown in Fig. 5, the MDO environment may solve the
section profiling problems with direct and inverse design methods.
With the inverse design method, the MDO environment first al-
lows engineers to specify desirable velocity and pressure distribu-
tions, and then computes discrete boundary points that meet the
requirements. Consequently, it generates section profiles by inter-
polating or approximating the control points that arise from aero-
dynamic computations. Obviously, the engineering efficiency
largely depends on engineer’s knowledge of optimal velocity and
pressure distributions, and benchmark data of previous designs.
Whereas, with the direct design method, the MDO environment
allows veteran engineers to manually adjust positions of control
points by mouse clicks and drags, and accordingly computes and
displays aerodynamic characteristics of the airfoil variants. Appar-
ently, this trial-and-error design method heavily depends on engi-
neering experiences and know-hows about optimal profile
curvatures.
Fig. 5. Modeling of blade section profile.
2D section profiles provide detailed geometric parameters like
inlet and outlet angles, stagger and wedge angles, thickness distri-
butions, diameters of leading and trailing edges.
In practice, about 5 or up to 7 basis section profiles are normally
created for a blade, in alignment with computation of S1/S2
streamsurfaces. These basis section profiles are further stacked
along a radial stacking line, which heavily influences the 3D airfoil
of a blade. Subsequently, 3D blade airfoils are generated by wrap-
ping up the stacked section profiles with boundary surfaces de-
fined by geometric interpolation or approximation algorithms. In
general, a blade surface should be generated in such a way that
the given section profiles are interpolated and each isoparametric
line of the turbine blade surface representation (between the given
profile curves) is a planar profile curve too and each profile planes
are parallel [22].
Blade streamsurface computations are very time-consuming
tasks, which are unlikely to attain optimal solutions in a few iter-
ations. Hence, prohibitive computation costs make engineers un-
able and reluctant to manually validate many blade geometric
variations for finding better flow characteristics. With recourse to
the MDO environment, engineers are able to easily change the
blade parameters such as inlet and exit angles, stagger angles,
chord lengths, and thickness distributions to verify a larger number
of flow simulations. And, sophisticated graphical representation
means make engineers feel very easy to validate and judge the
computation outcomes of aerodynamic parameters and geometric
parameters.
3.4. Flow characteristics analysis
Naturally, it is concerned if the desired aerodynamic perfor-
mance will be achieved or not with 3D blade airfoils resulted from
approximating surfaces that cover the skeletons of stacked basis
sections. Hence, 3D flow analysis has to be conducted to validate
 C. Wang / Advances in Engineering Software 68 (2014) 9–18
the aerodynamics performance, by using CFD tools. The 3D flow
analysis predicts flow characteristics through every turbine stage,
which includes mass flow rate, power, efficiency, velocities, pres-
sures, losses, etc.
3D flow analysis posterior to creation of 3D blade airfoils avoid
high computation costs of extensively running CFD systems in the
preliminary design stages. On the other hand, it allows engineers to
precisely assess blade structures on basis of aerodynamic simula-
tions. In a turbine, variations of geometric parameters exert con-
siderable influences on loss mechanisms like secondary loss, tip
clearance loss, and profile loss. Accordingly, the 3D flow analysis
provides engineers with explicit performance criteria for inces-
santly refining blade airfoils until losses are effectively decreased.
4. Blade airfoil modeling techniques
This section further discusses geometric modeling techniques
both for 2D and 3D blade airfoils in more detail. Even though
CAD systems are integrated in many collaborative design plat-
forms, there sound reasons of applying tailored geometric model-
ing techniques in aerodynamic blade design. Tailored geometric
modeling techniques constitute rudimentary parts of the MDO
environment.
In design of 2D blade profiles, geometries computed by the leg-
acy problem-solvers are represented as coordinates of points on
sectional surfaces, and stored in plain data files. It is very time con-
suming and tedious for engineers to manually process primitive
geometric data and regenerate 2D blade models with a commercial
CAD system. Apart from this drawback, commercial CAD systems
do not permit 3D blade airfoils to be constructed by stacking and
skinning 2D section profiles. Besides, using commercial CAD sys-
tems in aerodynamic blade design is quite likely to result in erro-
neous manipulation of data and lower work efficiency. Thus, it is
computationally impracticable for engineers to regenerate blade
geometric models in each aerodynamic computation by using a
CAD system.
Therefore, it is a common practice for aerodynamic engineers to
generate geometric blade models by using ad hoc geometric mod-
eling tools instead of complex commercial CAD systems. To meet
requirements of modeling blade airfoils, the MDO environment
provides a number of curve and surface generation techniques by
coding a number of polynomial interpolation and approximation
algorithms.
4.1. Geometric representations of 2D blade section profiles
As mentioned above, the quasi-3D blade engineering approach
first generates a number of blade cross-sections, and then stacks
these section profiles along a radial axis to form a 3D surface model
of a blade.
Invariantly, a 2D/3D blade airfoil is composed of four compo-
nents, i.e. the leading edge, suction, pressure, and trailing edge
curves/surfaces. Aerodynamics engineers apply the inverse analy-
sis algorithms to determine the desirable blade section profiles
that satisfying prescribed velocity and pressure distributions. The
profile curves are often controlled by a number of discrete control
points.
A 2D blade section profile determines the blade-to-blade ex-
haust flow through the turbine cascade. In this work, the control
points are either specified by global coordinates or by relative posi-
tions to the blade camber line. And, the four curve segments of a
planar blade profile can be generated by a number of piecewise
polynomial representations, such as simple cubic splines [5].
To avoid redundancy, this paper does not intend to depict the
mathematical equations of various polynomial integration and
15
approximation methods, which can be easily found in literature
like [23]. The MDO environment has comprehensively imple-
mented the commonly used curve interpolation/approximation
methods, which allows engineers to model blade profiles with flex-
ible combinations of curve representations.
First, the leading and trailing edges are represented by circular
arcs while the suction and pressure curves are interpolated with
cubic or higher order splines. Second, all the four curves are repre-
sented by B-splines or Bézier curves, a remarkable disadvantage of
which is their incompetency of accurately representing circular
arcs at the edges. Third, all the four curves are modeled by Non-
Uniform Rational B-Splines (NURBS), the merits of which are using
one uniform modeling method, and conveniently generating pre-
cise representations of circular leading and trailing edges.
After generating a blade section profile, engineers may interac-
tively adjust suction and pressure curves for meeting specified
blade thickness distributions, by numerically reconciling the geo-
metric discrepancies. In this way, the MDO environment provides
engineers with high flexibility of choosing pertinent ways to repre-
sent the blade profile curves. This enables engineers to efficiently
validate different blade profile options, in terms of geometric
parameters like stagger angle, inlet/outlet angles, solidity, thick-
ness distributions, etc. By integrating direct and inverse streamsur-
face solvers and geometric modeling techniques, the MDO
environment has created correlational mappings among blade sec-
tion profiles, flow losses, efficiencies, and other characteristics.
These correlational mappings make a change in one part automat-
ically trigger changes in other parts. Based on these sophisticated
correlational mappings, engineers can efficiently carry out a large
number of interactive emulations to evaluate mutual influences
between blade profiles and flow characteristics.
For instance, when engineer drag control points to designated
positions on screen by mouse clicks, the system automatically
regenerates the blade section profile and calculates consequential
flow properties. On the other hand, when engineers modify inputs
of aerodynamic performance, the system automatically calculates
the resulting changes in control point positions, and regenerates
blade section profiles.
4.2. Geometric representations of 3D blade airfoils
After building a number of section profiles, the system helps
engineers to create a solid blade airfoil by mounting up the profiles
along a stacking axis and closely skinning the stacked sectional
profiles. Obviously, there are many geometric parameters that con-
siderably influence the final blade shape, and hence its aerodynam-
ics performance. Unfortunately, there are no straightforward rules
that can guarantee optimal selections of profile stacking and sur-
face approximating parameters.
Apparently, the stacking and skinning processes are basically
based on engineering know-hows and historical inveteracies. To
provide engineers with a wide range of optional methods for pro-
file stacking and boundary skinning, the MDO environment formal-
izes and keeps a set of heuristics and know-hows extracted from
engineering experiences of veteran engineers, which allow guided
selections of the control parameters.
For instance, it is recommended that stator vane sections be
mounted up a stacking axis through the centers of the leading
edges or trailing edges, whereas, rotor blade profiles be mounted
up through their mass centers. This method of selecting stacking
locations is believed of being able to maintain high structural
integrity.
In practice, stacking axes are not always lines but frequently
curves, which permit bending and twisting effects of blade airfoils.
Therefore, apart from straight lines, the system permits stacking
axes to be optionally approximated with Bezier or NURBS curves,
16 C. Wang / Advances in Engineering Software 68 (2014) 9–18

which enable blades to lean or bend over in the circumferential or
axial directions.
And, the NURBS formulation is used to create 3D blade surfaces
because NURBS surfaces are generally believed of outperforming
other surface approximations. Fig. 6a presents an illustration of
stacked blade sections, Fig. 6b a 3D blade airfoil of NURBS surfaces,
and Fig. 6c an illustration of effects of stacking axes.
5. Flow characteristics analysis and design optimization
As noticed, design of blade airfoils are based on 1D and 2D aero-
dynamic computations, whereas, exhaust flows in axial turbines
are actually three dimensional. Consequentially, it is natural that
3D flow analysis has to be conducted, often by using commercial
CFD codes, to calculate exact velocity and pressure distributions
on the blade profiles, to predict various losses hence incurred,
and to precisely evaluate aerodynamic efficiencies of the blade
cascades.
Aerodynamic blade design is a multiobjective optimization
problem that tries to find a set of geometric parameters, which
simultaneously minimize a group of global and disciplinary objec-
tive functions, while satisfying a number of constraints. Mathemat-
ically, an aerodynamic blade design problem may be formulated as
below.
min F i ðXÞ i ¼ 1; 2; . . . ; m ð1Þ
subject to
8 work, and they produce turbulent secondary flows when mixing
< hl ðXÞ ¼ 0 l ¼ 1; 2; 3; . . . ;
> with the main flows. Maintaining appropriate tip clearances is a
g j ðXÞ  0 j ¼ 1; 2; 3; . . . ; ð2Þ
>
: 
xi  xi  xþi i ¼ 1; 2; 3; . . . ;
where X = (x1, x2, . . ., xn) e Rn are decision variables, Fi(X) are global
and disciplinary objective functions, hi(X) and gi(X) are equality and
inequality constraints.
Specifically, the objective functions in this work focus on blade
weights, aerodynamic efficiencies or losses due to secondary flows.
Weight reduction is simply formulated as reduction of blade vol-
umes. On the other hand, mechanisms and formulations of second-
ary flow losses are somewhat complex. Apart from the main axial
flows, there also exist a number of unsteady secondary flows in a
turbine, which are very heterogeneous in terms of directions and
velocity magnitudes. These secondary flows and their mixing with
the main flows result in a variety of losses.
Losses in a blade cascade are normally categorized into second-
ary flow loss, profile loss, tip clearance loss, endwall loss, shock
loss, etc. The term of secondary flow is generally used to address
flows that have non-zero velocity components normal to main
flows, which give rise to vortices and turbulences. It is widely ac-
cepted that losses caused by secondary flows constitute major part
of energy losses in turbines.
The profile loss results from energy dissipation on blade bound-
ary layers. The endwall loss is a kind of secondary flow losses result
from mixing vortexes and energy dissipation near hub boundary
layers.
The tip clearance loss results from flow leakage from pressure
surfaces through cleavages between blade tips and the engine case
to suction surfaces, due to pressure discrepancies. Tip leakage
losses are caused by the facts that leakage flows produce no useful
difficult task. Smaller tip clearances definitely decrease leakage
losses, however may not sufficiently allow for thermal growths
of blade heights, which is likely to result in severe abrasion and

Fig. 6. Blade section stacking and skinning.

 C. Wang / Advances in Engineering Software 68 (2014) 9–18
Fig. 7. Distributions of flow parameters along blade heights.
failure. Larger tip clearances guarantee safe operations but result in
considerable leakage losses.
Because of their crucial impacts on turbine efficiencies, a lot of
mechanisms or models have been developed to interpret and cal-
culate these losses. For instance, a classic formula for calculating
the overall pressure loss caused by secondary flows is given below
[24].
cos2 aout
Y s ¼ 4kðtan ain  tan aout Þ2 ð3Þ
cos am
where k is a parameter correlated with flow acceleration, ain, aout
and am are respectively flow incidence angle, exit angle, and aver-
age angle in the blade cascade.
Accordingly, a variety of loss models are implemented in the
system to provide engineers with tools for predicting losses com-
monly encountered in blade rows. For the sake of simplicity, we
do not replicate these loss models in this paper, which may be
found in previous literature, such as [24].
The system spontaneously displays flow characteristics in elab-
orate graphical representations, such as blade radial distributions
of flow exit angles, pressures, and energy losses, etc., as shown in
Fig. 7. The graphical representations of flow characteristics serve
as an indispensable booster to insightful comprehension of com-
plex flows through turbine stages.
The ultimate goal 3D flow analysis is to manifestly show dis-
crepancies between actual flows and desirable flows, which neces-
sitate continual refinements of blade airfoils.
17
6. Conclusions
This paper presents outcomes of the initial pilot stage of imple-
menting an MDO environment to aerodynamic blade design and
analysis. In this work, application and information integration
techniques are applied to build a shared cybernetic platform for
engineering computations. Then, sophisticated means of engineer-
ing flow management are developed to effectively coordinate the
design and analysis activities.
Technically, the aerodynamic blade design is implemented as
an iterative engineering process that is rudimentarily composed
of sub-processes of 1D meanline analysis, S1/S2 streamsurface
computations, 2D section profiling, 3D airfoil approximations,
and 3D flow analysis. Essentially, a number of curve/surface inter-
polation and approximation algorithms are coded to allow flexible
generation of 2D/3D blade airfoils. The tailored geometric model-
ing utilities save engineers from inefficiency and incompetency
of using commercial CAD systems.
Subsequently, the system enables engineers to efficiently con-
duct 3D flow analysis computations and to clearly comprehend
fluid behaviors against given blade airfoils. This makes blade air-
foils be iteratively improved until desirable aerodynamic perfor-
mance is met.
It is also recognized that aerodynamic blade design is a knowl-
edge intensive process where know-hows play an important role.
In this work, the intangible engineering know-hows are dealt in
two alternative ways. First, procedural rules are formally ab-
stracted and coded to support decision-making activities. Second,
a set of alternative approaches are provided in cases where no
18 C. Wang / Advances in Engineering Software 68 (2014) 9–18
explicit procedural rule can be formulated, which allow engineers
efficiently compare a number of design options.
Convincingly, the pilot application has yielded remarkable tech-
nical benefits that have greatly lessened suspicion about transfer-
ring conventional aerodynamic blade design to the new
cybernetic platform. The fruitful outcomes of the pilot stage have
effectively bolstered continual implementations of heat transfer
analysis, structural analysis, and multidisciplinary design optimi-
zation in design of blades, discs, and other components. And, the
extended technical applications of the MDO environment are to
be addressed in following up papers.
Acknowledgments
The author would thank his colleagues and postgraduate stu-
dents who have taken part in the projects concerned with this
work for their strenuous efforts to develop and implement the sys-
tem. The author is also grateful to the aerodynamics engineers who
considerably contribute to successful application of the computer-
ized techniques. Finally, the author thanks the two anonymous
reviewers for their constructive comments and suggestions for
improving the paper.
References
[1] Talya SS, Rajadas JN, Chattopadhyay A. Multidisciplinary optimization for gas
turbine airfoil design. Inverse Problems Eng 2000;8:283–308.
[2] Joubert H, Quiniou H, Turbomachinery design used intensive CFD. 22nd
Congress of International Council of the Aeronautical Sciences. 28 August – 1st
September, 2000, Harrogate, UK.
[3] Wang C, Zhang Y, Song G, Yin C, Chu C. An integration architecture for process
manufacturing systems. Int J Comput Integr Manuf 2002;15:413–26.
[4] He W, Xu L. Integration of distributed enterprise applications: a survey, IEEE
Transactions on Industrial Informatics, 2013 http://dx.doi.org/10.1109/
TII.2012.2189221.
[5] Talya SS, Chattopadhya A, Rajadas JN. An integrated multidisciplinary design
optimization procedure for cooled gas turbine blades. In: 41st AIAA/ASME/
ASCE/AHS/ASG Structures, Structural Dynamics, and Materials Conference and
Exhibit, Atlanta, GA, USA, 3-6 April 2000.
[6] Houstis EN, Catlin AC, Tsompanopoulou P, et al. GasTurbnLab: a
multidisciplinary problem solving environment for gas turbine engine design
on a network of nonhomogeneous machines. J Comput Appl Math
2002;149:83–100.
[7] Koini GN, Sarakinos SS, Nikolos IK. A software tool for parametric design of
turbomachinery blades. Adv Eng Software 2009;40:41–51.
[8] Qiu X, Anderson M, Japikse D. An integrated design system for
turbomachinery. J Hydrodynamics 2010;22:358–65.
[9] Wang C. Insights into developing an MDO environment for complex
engineering systems. Journal of Engineering Design. Submitted for review.
[10] Wang C. Multidisciplinary engineering application integration based on
component technology. In: Maropoulos PG, Newman Stephen T editors.
Proceedings of the 4th international conference on digital enterprise
technology. Bath UK. 19–21 September, 2007; p. 185–92
[11] Wang C, Xu L. Parameter mapping and data transformation for engineering
application integration. Inf Sys Frontier 2008;10:589–600.
[12] Wang C. Advances in information integration infrastructures supporting
multidisciplinary design optimization. Enterprise Inf Sys 2012;6:265.
[13] Feng G, Cui D, Wang C, Yu J. Integrated data management in complex product
collaborative design. Comput Ind 2009;60:48–63.
[14] Huang Z, Wang C, Ma M. A robust archived differential evolution algorithm for
global optimization problems. J Comput 2009;4:160–7.
[15] Huang Z, Wang C. Optimal design of aeroengine turbine disc based on kriging
surrogate models. Comput Struct 2011;89:27–37.
[16] Yu J, Cha J, Lu Y, Xu W, Sobolewski M. A CAE-integrated distributed
collaborative design system for finite element analysis of complex product
based on SOOA. Adv Eng Software 2010;41(4):590–603.
[17] Xu LD, Viriyasitavat W, Ruchikachorn P, Martin A. Using propositional logic for
requirements verification of service workflow. IEEE Trans Ind Inf
2012;6:639–46.
[18] Lwin T, Vu NA, Lee JW, Kim S. A distributed web-based framework for
helicopter rotor blade design. Adv Eng Software 2012;53(November):14–22.
[19] Tournier JM, El-Genk MS. Axial flow, multi-stage turbine and compressor
models. Energy Convers Manage 2010;51:16–29.
[20] Denton JD, Dawes WN. Computational fluid dynamics for turbomachinery
design. Proc Inst Mech Eng J Mech Eng Sci. Part C 1999:107–204.
[21] Ristic M, Brujic D, Ainsworth I. Measurement-based updating of turbine blade
CAD models: a case study. Int J Comp Integr Manuf 2004;17(4):352–63.
[22] Hoschek J, Muller R. Turbine blade design by lofted B-spline surfaces. J comput
Appl Math 2000;119:235–48.
[23] Hearn D, Baker MP, Carithers WR. Computer Graphics with OpenGL. 4th ed.
Pearson Education Asia Ltd. and Publishing House of, Electronics Industry;
2012.
[24] Wei N. Significance of loss models in aerothermodynamic simulation for axial
Turbines. Doctoral Thesis, Department of Energy Technology, Division of Heat
and Power Technology, Royal Institute of Technology, Sweden, 2000.