You are on page 1of 0

1

CHAPTER 1
INTRODUCTION
1.1 Overview
Computational Fluid Dynamics (CFD) has progressed rapidly as a discipline
and is being increasingly used to complement the wind tunnel measurements. Prior to
1960, the primary tool for the analysis and development of aerodynamic
configurations was the wind tunnel measurements and empirical techniques of
analyses. In wind tunnel measurements, the real flow conditions are simulated by
using the similarity principles. Since it is almost impossible to test the full-scale
vehicle in the wind tunnel due to size constraints, usually a smaller scale model is
tested, for which the similarity of Reynolds and Mach numbers is a major issue. The
Reynolds and Mach numbers of the free flight conditions must be matched in wind
tunnel measurements in order to simulate the actual flow conditions as much as
possible. In cases where these similarities could not be attained, the flow field
matching is achieved by various modifications on the model surface, such as a trip
wire or applying surface roughness on the model. Wind tunnel experiments are
expensive due to the preparation of the small scale instrumented model and the cost
of run times of the tunnels. More than twenty thousand hours of wind tunnel testing
were done in the development of some modern aircraft, such as the F111 and Boeing
747 [1]. However, once the infrastructure of the experiments is prepared, it becomes
straightforward to get systematic measurements for the aerodynamic behavior of the
vehicle at various flight configurations. A model can be tested and modified in the
2
light of pressure and force measurements and flow visualizations performed on the
model. Wind tunnel tests are often limited by instrumentation constraints, precise
model manufacturing, tunnel calibration, flow quality, wall and support interferences
and aeroelastic effects. CFD methods are being developed as an alternative to wind
tunnel experiments, to replace the real experiments by the so-called numerical
experiments. Compared to wind tunnel tests, CFD methods are less expensive and
require less time. The accuracy of the CFD solutions depends upon the complexity of
the physical and numerical modeling utilized.
Starting nearly thirty years ago and becoming increasingly important in the last
decade, CFD has become a powerful tool in the analysis and understanding of any
type of flow phenomena including inviscid/viscous, compressible/ incompressible,
heat and mass transfer, phase change, chemical reaction, etc. about any kind of
geometry; such as aircraft, automobile, ship, building, pump, fin, etc. CFD has a
wide range of engineering applications not only in aerodynamic fields, but also in a
broad variety of fields such as automotive, engine and ship industries, manufacturing
engineering, chemical engineering, civil engineering, oceanography, meteorology,
astrophysics, nuclear science and even in medicine for studying blood flow. The
ultimate goals of CFD include; vehicle design, performance analysis, definition of
loads for structural design, aeroelastic analysis including flutter and provide the
mathematical model for the full flight simulation. Although, the current capability of
CFD does not allow to accurately satisfy all these goals, improvements in the
numerical methods and the computer performances increase its applicability in the
overall design phase. Today, there is no doubt that CFD and wind tunnel data are
complementary to each other to allow engineers to investigate problems and assess
their designs.
The simulation of fluid flow starts with a physical model of the flowfield. The
physical system is then represented with a suitable mathematical model. The choice
of an appropriate, and usually simplified, mathematical model represents a tradeoff
between the accuracy required in the simulation and the computational complexity
3
and cost. Next step is the application of a numerical technique to compute the
solution of the selected mathematical model. Physical modeling or domain
discretization can be achieved by dividing the flow field into small control elements
to each of which the conservative governing flow equations are applied in a
discretized manner. These elements can be generated either in a structured or
unstructured manner. Selection of a numerical scheme to be applied for equation
discretization is of course a function of the elements used. For example, finite
difference method (FDM) requires highly regular elements (structured grids),
whereas the finite volume method (FVM) has a great flexibility on the type of
elements used (unstructured grids), thus applicable to more complex geometry and
problems. The development of the finite volume and unstructured grid generation
techniques has broadened the capabilities and the range of applications of the CFD
compared to the traditional finite difference and the structured-grid methods. The
unstructured grid approach offers some significant advantages for simulating flows
over complex geometries. This methodology now enables engineers to generate
tetrahedral grids about complex configurations in a relatively short time interval [2].
Inviscid flow solutions are readily obtained by using a number of well-proven
unstructured flow solvers [3]. Although the current research in CFD is focused upon
the NavierStokes equations, Euler equations can still provide fairly accurate flow
field simulations throughout much of operating envelope of an aircraft. Because of
their ability to capture both strong and vortical flows, Euler simulations have become
a standard tool for current aircraft design processes. In the design of the YF-23, an
Euler analysis was used quite successfully and inviscid codes have proved to be of
significant in aerodynamic design [4]. Recent advances and trends in CFD [5-7] have
made the use of unstructured grids in three-dimensional computations about complex
high-lift aerodynamic configurations more feasible [6].
4
1.2 Thesis Scope and Outline
The purpose of this thesis is to apply a CFD tool, CFD-FASTRAN [8], to solve
inviscid, compressible, subsonic flow over a medium range cargo aircraft in cruise
and in high-lift (landing) configurations.
In order to assess the performance and check various parameters of the flow
solver, simple wing configurations are analyzed at first. Then the solution of full
aircraft geometry is performed. In the cruise configuration study; the fuselage, wing,
nacelle, gondola, horizontal and vertical tails are modeled. In the so-called landing
configuration study, inboard and outboard wings with single slotted flaps and flap
hinge fairings are also modeled. In both cases, engine intakes are modeled as closed
surfaces.
Chapter 1 of the thesis gives a brief description of the CFD methodology used,
starting from model set-up to the final solution. The geometry and the flow solver are
introduced. In addition, a literature survey on CFD calculations is presented.
Chapter 2 is about grid generation. A brief explanation of CFDRCs grid
generation code, CFD-GEOM [9], is presented and the principles of the unstructured
grid generation process are explained. For computing flows over complex
geometries, the use of unstructured grids offers considerable savings in the number
of grid points and therefore a reduction in the grid generation time. CFD-GEOM has
an automatic and controllable volume tetrahedral mesher for this purpose. Several
examples of complex grids generated by using CFD-GEOM are also presented in this
chapter.
Chapter 3 describes the features of the flow solver. The capabilities of the
CFD-FASTRAN flow solver and various computational methods applied are
discussed. Governing differential equations and the finite volume method are stated
here.
5
Chapter 4 discusses several test cases in 2-D and 3-D. For the 2-D test case, a
NACA 0012 symmetric airfoil is used and the results are compared with the
experimental data. For the 3-D test case, the ONERA-M6 transonic wing is used. For
this test case, surface grid quality assessment is performed by using three different
surface meshes and the results are compared to one another and to the solutions of
another code. Several test cases on the clean-wing (wing alone) are considered for
grid convergence studies and solver parameter checks. A detailed description of each
configuration is given in Chapter 5. Both wing and aircraft configurations are solved
at cruise and landing conditions. The computational results for all the cases studied
are also presented in this chapter.
Finally, Chapter 6 presents conclusions that can be drawn from the present
study. Suggested future work is also included to extend and improve the applicability
of the code.
1.3 CFD Method (from CAD Model to Solution)
Today, Computer-Aided Design (CAD) systems can typically represent the
surfaces of aerodynamic vehicles with a set of parametric surfaces such as Non-
Uniform Rational B-Splines (NURBS) [10]. NURBS can be used to represent most
spline and implicit curves without loss of accuracy [11]. However, when data is
exchanged, the CAD models usually lose their precision/integrity and produce
surface representations which, in some cases, are severely distorted and difficult to
work with [12]. A more important step is that the physical model should be correctly
represented by a CAD package. In this work, the aircraft model has been constructed
using the I-DEAS CAD software and the model is imported into the CFD-GEOM
grid generator in the Initial Graphics Exchange Specifications (IGES) format.
However, this transformation introduces some small problems, such as untrimmed
surfaces and duplications. Several imperfections and curve anomalies actually result
from inadequately defined CAD geometries that are not suitable for grid generation.
These must be fixed before the actual grid generation process is started.
6
Numerical solution of the problem is only a small part of the entire analysis
(Figure 1.1). Before any numerical solution can be carried out, the physical domain
must be represented with a computational grid. Selection of an appropriate physical
modeling is a priority before grid generation. Judgement has to be made to select the
solution method. Then, from the solver, output data is post-processed to extract
meaningful information. Interpretations of the solutions are also essential to provide
and deduce the final results for the physical problem.
Figure 1.1 Typical Steps in Computational Fluid Dynamics.
1.4 Geometry
The geometry used in this work pertains to a conventional type, medium range
cargo aircraft (Figure 1.2). It has cantilever high-wing monoplane and raked wing-
tips. The fuselage has a flattened circular cross section and is upswept at rear. The
horizontal tail is unswept with a fixed negative incidence. The vertical tail has a
sweptback fin. The model assumes no aileron, elevator or rudder deflections. The
landing gear and propeller are also omitted but the gondola and the engine nacelle
(with closed intake) are modeled.
For computational economy, a semi-span model of the geometry is considered
for the studies. The model has two distinct wing configurations; flaps deflected and
flaps retracted. Flaps and their fairings cause additional complexity and an increase
7
in the number of surfaces because of sharp corners and extremely narrow gaps.
Before the full analysis of the aircraft configuration, simple wing-only (clean-wing)
cases are solved. The left wing of the aircraft is separated from the fuselage and
extruded up to the aircraft symmetry plane. The clean-wing serves as a convenient
case to test the grid generator and the flow solver. Figure 1.2 shows various views of
the aircraft drawn in I-DEAS software and Table 1.1 gives several features of the
model that are specified in ref. [13].
Figure 1.2 Three views of the medium range cargo aircraft.
Table 1.1 Features of the model.
Length overall 21.353 m
Fuselage max. depth 2.615 m
Span 25.81 m
Chord 3.0 m
Wing mean chord 2.62 m
Wings gross area 59.1 m
2
Flap deflection angle (landing) 23
Wing airfoil NACA 65
3
-218
Horizontal and vertical tail airfoils NACA 64
1
-012
8
1.5 Flow Solver
There are many available CFD solvers for commercial and research
applications (see ref. web site: 14). Each flow solver has its own specific features
that make it applicable to certain types of flows and geometries. These features are
mostly related to the grid systems, flow types and conditions, time integration and
spatial discretization algorithms, turbulence models and visualization capabilities.
Many FORTRAN based CFD codes dedicated to special problems are freeware
(noncommercial), but most in two dimensions.
Code development is a long process and even the initial validation phase of a
newly developed algorithm may take a year or two for a CFD researcher [15].
Considering the time limitations and the complexity of the geometry, selection of a
commercially available code is inevitable. However, using an available software is
not trouble free and has its own problems. As Whitfield states The codes do not
produce results, but people produce results using the codes [15].
Amongst the available commercial and public codes, CFD-FASTRAN
provides highly accurate simulations of compressible flow and incorporates robust
higher-order accuracy numerics [16]. It supports various grid systems including
multi-block-structured grids, unstructured tetrahedral grids, chimera/overset grids
and hybrid grids. A full range of flow conditions based on the Euler/Navier-Stokes
equations is handled by CFD-FASTRAN [16]. In the present work, steady, inviscid,
compressible module of the code is used with cell-centered, density based, finite
volume upwind solver options.
9
1.6 Literature Survey
The history of numerical techniques for the solution of the inviscid Euler
equations goes back to the early 1950s, with the first-order methods of Courant et al.
[17]. The milestone for the modern development of numerical schemes for time-
dependent Euler equations is the pioneering work of Lax and Wendroff [18]. For
steady-state calculations, implicit generalizations of the Lax- Wendroff schemes have
been developed by Lerat [19].
Parallel to the development of space-centered schemes, alternative
discretization methods have also been developed which relate to the physical
propagation properties of the solutions of the Euler equations. These non-space-
centered schemes called as upwind schemes and the first explicit scheme was
introduced by Courant et al. [17]. Upwind schemes have become very popular over
the last decade. Several extensions to second-order accuracy and implicit time
integrations have been developed such as flux vector splitting (FVS) methods of
Steger and Warming [20] and Van Leer [21]. The second group of schemes, known
as flux difference splitting (FDS) methods is approximate Riemann solvers of
Engquist and Osher [22] and Roe [23]. These methods are in line with Godunovs
[24] method which solves the locally one-dimensional Euler equations for
discontinuous neighbouring states (the Riemann problem) [25].
The finite volume spatial discretization was introduced into the field
independently by McDonald [26] and MacCormack and Paullay [27] for the solution
of two-dimensional, time-dependent Euler equations and extended by Rizzi and
Inouye [28] to three-dimensional flows. The finite volume method takes the full
advantage of using arbitrary control volumes. The concept of using arbitrary control
volumes to solve numerically the conservation laws was established by the late 70s,
at least in theory [29]. Jameson and Mavriplis [30] reported some of the earliest
solutions of the two-dimensional Euler equations on regular triangular grids by
subdividing quadrilateral grids. They extended finite volume schemes on structured
10
grids to triangular grids in a cell-centered setting. In 1986, Jameson et al. [31]
presented a paper dealing with solutions of inviscid transonic flow over a complete
aircraft. The papers contributions included grid generation for complex geometries
using the Delaunay triangulation approach and the development of a cell-vertex
based algorithm [3].
Frink et al. [5, 32] developed an upwind cell-centered three-dimensional flow
solver. They employed a weighted averaging procedure that interpolated variables
from the centers of tetrahedra to the vertices and used these vertex values to compute
the gradients within each tetrahedron. Implicit cell-centered upwind algorithms for
three-dimensional unstructured meshes were reported by Batina [33].
Several successful applications of unstructured, finite volume, Euler solutions
have been reported in the last decade [12,34-37]. Finite volume Euler flow
simulations of an unstructured code (Airplane) of Jameson on complex geometries
were performed at Dornier using an Airbus A320 wing body and complete fighter
aircraft configurations [34].
Another unstructured Euler code application for subsonic analysis of F-16XL
model is stated by Lessard in [35]. Grid quality effects on the solutions were studied
on the ONERA M6 wing using another unstructured Euler code by Frink et al. [5].
Similar work to the present study was performed on NASA Langley Transport
Systems Research Vehicle (TSRV) aircraft (Boeing 737-100) by Dodbele [36]. Two
computational methods, a surface panel method (VSAERO) and an Euler method
(USM3D) employing unstructured grid methodology, were used to analyze a
subsonic transport aircraft in cruise and high-lift conditions. Grid generator
(VGRID3D) and flow solver (USM3D) use similar algorithms to CFD-GEOM and
CFD-FASTRAN, respectively. Dodbeles work [36] was one of the first applications
to develop unstructured grids on a 3-D high-lift aircraft configuration and to apply
Euler calculations on a high-lift case in subsonic flow using unstructured grid.
Sample grids from this work are given in Figure 1.3.
11

Figure 1.3 Unstructured Surface Grids of TSRV in Cruise and Landing
Configurations [36].
A more detailed work on unstructured grid generation for complex 3-D high-
lift configurations was examined by Pirzadeh [12]. Several difficulties encountered
during modeling, geometry definition and grid generation on complex parts were
also stated in his presentation.

You might also like