Sei sulla pagina 1di 6

Department of Mechanical & Aerospace Engineering

CARLETON UNIVERSITY

AERO 4304: Computational Fluid Dynamics


Winter 2013

Lecture 1: Introduction
1

Lecture summary

This is the first lecture of the AERO4302 course in Winter 2013 term. In this lecture,
we will discuss:
A review of the course syllabus
An overview of the course assignments and project
What is CFD?
Why perform a CFD analysis?

Review of the course syllabus

AERO 4302 is an undergraduate computational fluid dynamics (CFD) course. The


course outline that is distributed on the first day of classes provides a brief introduction to the course. The objective of the course as outlined in the syllabus is for
students to acquire a sufficient knowledge of the principles of CFD to enable them to
be knowledgeable users of commercial CFD software, able to make informed decisions
on the setup and running of a CFD simulation, and able to assess the accuracy and
applicability of the results.
Detailed information on the instructor, the course requirements, reference material, and a tentative list of topics to be covered are presented in the syllabus. Note:
the list of course topics in the syllabus is a tentative list; the topics may change or
be presented in a different order within the term.

Review of the course assignments and project

As with anything, learning CFD requires practice, and there is no teacher better that
your own experience. Therefore, I aim to make this course as hands-on as possible.
The nature of the course makes this difficult; the content of CFD is frequently quite
abstract. Practice with the course material will come in the following ways: (i) simple
R
programming assignments; (ii) simulating simple canonical flows in ANSYS CFX ;

AERO 4304

Lecture 1

and (iii) performing a simulation (or series of simulations) of a somewhat larger scope
as a course project.

3.1

Programming assignments

As the name implies, computational fluid dynamics requires an understanding of both


the fluid-dynamic principles that govern the flow being simulated, and the computational approaches that are used to perform the simulation. To aid students in
developing such an understanding, you will be required to apply the numerical methods discussed in the course to solve simple fluids problems. These assignments will
involve writing your own computer code in your preferred programming language. For
students with minimal programming experience, dont fret; the programming effort
will not be huge, and I will provide enough help that you will be able to focus on the
details of the numerical method rather than the mechanics of computer programming.
That said, I think that it is invaluable for any student of CFD to spend time in front
of a computer agonizing over why your code wont work. That experience helps you
to become familiar with the nuts and bolts of CFD, it allows you to personally experience the strengths/weaknesses of different methods, and it helps you develop an
appreciation for the limitations and inaccuracies of numerical methods. Without personal experience implementing the numerical methods discussed in this course, there
is a danger that you may treat commercial CFD solvers as magical black boxes and
treat the results they spit out as gospel.

3.2

Simulation of canonical flows in CFX

In the same spirit as the above statements, an understanding of the capabilities and
limitations of commercial CFD solvers (such as ANSYS CFX) is important in order
to assess the accuracy of the results obtained from such solvers. The intuitive userinterface, the ease by which CFX allows a user to define and a run a simulation, and
the colourful output masks the fact that deliberate, informed choices on the part of
the user are necessary in order to obtain meaningful results. The best way to learn
this is to experience first-hand how CFX handles simple canonical flows, particularly
those with analytical solutions that allow comparison against the computed results.

3.3

Course project

Ultimately, the goal of learning CFD is to be able to setup and solve a problem that
is of interest to the learner. The best way to do this is through a project where you
are free to make your own choices on how you will setup the geometry, boundary
conditions, and simulation method. A list of possible topics will be provided, or
students can propose their own. The selected topic must be approved in advance by
me, and each topic must have experimental or high-quality numerical data available

AERO 4304

Lecture 1

to compare your results against. The detailed guidelines regarding the project are in
the project handout distributed with the course syllabus.

What is CFD?

CFD is the science of predicting the behaviour of fluid flow through the solution
of the equations of motion that govern the flow. While these governing equations
(called the Navier-Stokes equations) have been known for over a century, only a few
analytical solutions exist for a limited number of flows. These known solutions are
often helpful in understanding more complicated cases, but they do not provide us
with much meat to chew when it comes to engineering analysis or design. CFD
addresses the problem by converting the governing equations from their original form
(that is, continuous partial differential equations, or PDEs; these are discussed more
in Lecture 2) into algebraic equations that can be solved using numerical methods
computed with an electronic computer.
This process of converting continuous PDEs into algebraic equationsreferred
to as discretizationand then solving them with a computer is what CFD is, in a
nutshell. The overall process is shown in Fig. 1. Starting at step (1), the governing
equations continuously govern the behaviour of a flow variable (say, a function of
interest f plotted as a continuous curve) such that it varies continuously at all locations in space and time. The continuous PDE governing f is discretized in step (2)
to obtain an approximation to f applied at discrete points in space or time. At step
(3), the approximate equations at each discrete location within a domain of interest
are collected into a system of algebraic equations. To constrain the system, boundary
conditions and/or initial conditions need to be defined in step (4). The constrained
system of discrete algebraic equations that result (step 5) can then be solved using
a computer in step (6). The results produced by the solution are analyzed in step
(7), where the choice of governing equations, the discretization process, and the form
of the constraints can be modified if needed, and the process is repeated until the
results give the user what s/he needs.
Figure 1 indicates that CFD covers several fields, including calculus (to understand
the governing equations), fluid mechanics (to understand what the equations mean
physically and how they can be simplified), numerical methods (to select or develop
the appropriate discretization method and to solve it numerically), and computer
science (to execute the solution method using the hardware within the computer).
Thus, CFD is highly multidisciplinary; it straddles multiple branches of science. This
idea is shown in Fig. 2(a), where CFD lies at the intersection of fluid mechanics,
mathematics, and computer science. Beyond straddling these branches of science,
CFD also makes use of multiple methods of analysis, as seen in Fig. 2(b). Of course,
the goal of CFD is a simulation, but the simulation must have a strong foundation in
the theory in order to produce accurate results. And how does one know if the results
are accurate? Ideally, through comparison with experiments. Thus, CFD requires

AERO 4304

Lecture 1

Figure 1: Schematic illustrating the overall process of CFD.


the appropriate combination of theory, simulation, and experimental measurements,
and practitioners need to be sufficiently knowledgable in all these areas.

Why perform a CFD analysis?

Imagine that you are designing a race car and you want to optimize the shape of the
front-end of the car. Obviously, you cant build dozens of different geometries and
take each out onto a racetrack and choose the one thats fastest, so you decide to
perform some experiments on scale models. But what size of model do you choose?
You recall from your Fluids II course that for many flows, non-dimensionalization
of the Navier-Stokes equations leaves only the Reynolds number as the independent
parameter, and based on this fact, you build a 1:10 scale model and choose a wind
tunnel with a velocity 10 times the speed of the race car, say 1000 km/hr or about
280 m/s. But you quickly realize that achieving this speed in a wind tunnel requires
huge amounts of power, and furthermore, now the Mach number scaling of your car
is way off. You decide to ignore the Reynolds number scaling and just use 100 km/hr.
You build your scale model, place it in the wind tunnel, and turn the flow on. But
quickly, you realize that the experiment is missing the effect of the ground moving
beneath the stationary scale model of the car. You decide to build a conveyor belt
below the model and move it at the same speed as the oncoming air. By this time,
so much time and money has been expended that your company has gone bankrupt.

AERO 4304

Lecture 1

Figure 2: CFD is a highly multi-disciplinary subject, involving several (a) branches


of science and (b) methods of analysis.
This silly story illustrates some of the difficulties that are encountered in experiments that are not too difficult in CFD, such holding multiple scaling parameters
constant or implementing a range of boundary conditions. Sometimes the flow behaviour in an experiment may be modified by presence of the measuring apparatus.
Sometimes a experiment may give a good measure of the global behaviour of the flow
(say, the drag of your race car), but little detailed information in one part of the flow,
like if flow separation occurs on a surface or if the wall temperature exceeds a threshold value. Certain flow properties are difficult or impossible to measure accurately,
particularly in turbulent flows. Sometimes you want to learn about a process that
every effort is made to prevent, such as explosions or forest fires. In these instances,
the benefits of CFD become apparent. Although CFD can never completely replace
experimentation (for reasons that will become evident in this course), it has a number
of distinct advantages:
It can often be produced inexpensively and quickly. Computing costs have been
falling steadily for the past 50 years, and this trend appears to continue into
the foreseeable future.
It generates complete information of the flowfield. Detailed and complete information of the flow under observation can be obtained and studied.
It allows easy and quick configuration changes. This allows parametric studies
over wide ranges of parameters.
It has the ability to simulate ideal conditions. Simplifications that are not
possible in the real world can be implemented, allowing one to focus attention
on a few essential parameters and eliminate irrelevant features.
These and other benefits have led to the proliferation of CFD in a wide variety
of industries, including aeronautics, energy, health, weather prediction, and motor

AERO 4304

Lecture 1

sports. Type computational fluid dynamics into an online image search engine like
Google and youll be amazed by the wide variety of flows that have been simulated.
Of course, CFD calculations are never completely exact, and performing a CFD
simulation opens the door for many other sources of error to be introduced into the
analysis. Therefore, an important focus of this course is see how errors can occur in
each step of the CFD process and how they can be minimized.

Potrebbero piacerti anche