Theory and Numerics for Problems of
Fluid Dynamics
COVID19 version
Miloslav Feistauer
‘Charles University Prague
Charles University Prague, Faculty of Mathematics and Physies
2020PREFACE,
‘This text should serve as a source for the course “Theory and Numeries for
Problems of Fluid Dynamics”, delivered at RWTH Aachen in April/May 2006.
‘The main purpose of this course is to give a survey on the theory of ineompress-
ible Navier-Stokes equations. We also discuss the finite element method for the
numerical solution of viscous incompressible flow. Moreover, we are concerned
with some results in the theoretical and numerical analysis of compressible flow
More details can be found in the books
R. Temam: Navier-Stokes Equations. Theory and Numerical Analysis, North-
Holland, Amsterdam, 1977.
\V. Girault, P-A. Raviart: Finite Element Mathods for Navier-Stokes Equations.
Theory and Algorithms. Springer, Berlin, 1986.
M. Feistauer: Mathematical Methods in Fluid Dynamics. Longman Scientific &
Technical, Harlow, 1993.
M. Feistauer, J. Feleman, I. Straskeaba: Mathematical and Computational Meth-
cds for Compressible Flow, Clarendon Press, Oxford, 2004, ISBN 0 19 850588
41
CONTENTS
Basic Equations
1.1 Governing equations
11.1 Deseription of the flow
1.1.2. The continuity equation
1.1.3. ‘The equations of motion
1.1.4 The law of conservation of the moment of mo-
mentum
1.1.5. The Navier-Stokes equations
1.2. Incompressible flow
Mathematical theory of viscous incompressible flow
2.1 Funetion spaces and auxiliary results
22 The stationary Stokes probier
2.2.1 Stokes problem with homogeneous boundary con-
ditions
2.2.2 Stokes problem with nonhomogeneous bound-
ary conditions
2.3. The stationary Navier-Stokes equations
23.1 The Navier-Stokes problem with nonhomoge-
neous boundary conditions.
24 The Oseen problem
2.5 Functions with values in Banach spaces
2.5.1 Lebesgue Spaces of Functions with Values in a
Banach Space.
2.6 The nonstationary Navier-Stokes equations
2.7 Solvability of the nonstationary problem
27.1 Uniqueness
Finite element solution of incompressible viscous flow
3.1 Continuous problem
BALL Stokes problem
3.1.2 Navier-Stokes problem
3.2 Discrete problem
32.1 Diserete Stokes problem
3.3 Choice of the finite element spaces
34 Babutka Brezzi condition
34.1 Verification of the validity of the BB condition
34.2 Examples
3.5. Existence of an approximate solution
3.6 Brror estimates
10
2
18
20
2
2
2
7
30
32
32vii CONTENTS.
36.1 Abstract error estimate for the velocity
3.6.2 Abstract error estimate for the pressure
3.7. Numerical realization of the diserete problem
3.7.1 Solution of the saddle point system
38 Discrete Navier-Stokes problem
38.1 Iterative processes
3.8.2. Diseretization of the nonstationary Navier-Stokes
problem
References
39
40
4l
4
4
431
BASIC EQUATIONS
Problems of fnid dynamics play an important role in many areas of science and
technology. Let us mention, e.g.
‘airplane industry,
mechanical engineering,
turbomachninery,
ship industry,
civil engineering,
chemical engineering,
food industy,
environmental protection,
meteorology,
oceanology,
medicine,
‘The image of the flow can be obtained with the aid of
1) experiments (e.g. in wind tunnels), which are expensive and lengthy, some
times impossible,
'b) mathematical models and their realization with the use of numerical meth=
‘ods on modern computers. The numerical simulation of flow problems constitutes
the Computational Fluid Dynamics (CFD). Its goal is to abtain results compa-
rable with measurements in wind tunnels and to replace expensive and lengthy
experiments.
1.1 Governing equations
Let (0,7) C IR be a time interval, during which we follow the fluid motion,
and let @ CR", N = 1,2,3, denote the domain occupied by the fluid. (For
simplicity we assume that it is independent of time t.)
LL Description of the flow
There are two possibilities for describing the uid motion: Lagrangian and Bue
lerian,
‘We shall use here the Bulerian description based on the determination of the
velocity v(at) = (0)(2,t),---» vw(a,t)) of the fluid particle passing through the
point # at time t
‘We shall use the following basic notation: p ~ fluid density, p ~ pressure, @
absolute temperature. These quantities are called state variables.
Tn what follows, we shall introduce the mathematical formulation of funda
‘mental physical laws: the law of conservation of mass, the law of conservationa BASIC EQUATIONS
of momentum and the law of conservation of energy, called in brief conservation
laws.
1.12 The continuity equation
2 .t)+civ(ola,tvee,9)=0, CE (0.7), 20, (aa)
is the differential form ofthe law of conservation of mass.
11.3 The equations of motion
Basic dynamical equations equivalent with the law of conservation of momentum
hhave the form
oy
on;
Joon + ais (ra) = wht (112)
This can be written as
2 (pw) + de (p80) = pf + av (11a)
Here v9 is the tensor with components ¥.2).i, +N, 7 are components
of the stress tensor T and J; are components of the density of the outer volume
force f.
Ld The law of consereation of the moment of momentum
Theorem 1.1 The law of conservation of the moment of momentum is valid if
and only if the stress tensor T is symmetric.
11.5 The Navier-Stokes equations
‘The relations between the stress tensor and other quantities describing fluid flow,
particularly the velocity and its derivatives, represent the so-called rheological
‘equations of the fluid. The simplest rheological equation
pl, (ua)
T=
characterizes inviscid fluid. Here p is the pressure and I is the unit tensor:
1,0,0
t=(0.1,0) for
0,0,1
Besides the pressure forces, the friction shear forces also act in real fuids as a
consequence of the viscosity. Therefore, in the case of viscous fluid, we add a
contribution 7’ characterizing the shear stress to the term —pl:
(1s)
T=-pl+T’. (1.2.6)INCOMPRESSIBLE FLOW
On the basis of the so-called Stokes” postulates, it is possible to derive the de-
pendence of the stress tensor on the thermodynamic variables and the velocity
deformation tensor D(®) = (4is(®))3 jay with
dy(v
x (uaz)
‘Theorem 1.2 The stress tensor has the form
T= (-p+ Adivn) 1+ 2uD(0), (113)
where A, are constants or scalar functions of thermodynamical quantities.
If the stress tensor depends linearly on the velocity deformation tensor as in
(1.1.8), the fuid is ealled Newtonian,
We get the so-called Navier-Stokes equations
M2) avipo 00) ans
= pf — gradp + grad(A div v) + div(2pD(v)).
Here p> 0 and \ are called the first and the second viscosity coefficients,
respectively, is also called dynamical viscosity.
1.2 Incompressible flow
We divide fluids in liquids and gases. Gases, also called compressible fluids, have
variable density, whereas liquids, called incompressible fluids, have a constant
density p= const. > 0. For incompressible fuids, n the continuity equation and
the Navier-Stokes equations can be written in the form
div o a2)
(1.2.2)
Moreover, assuming that 1 = const. > 0 and denoting
viscosity), Navier-Stokes equation (1.2.2) reads
lp (=kinematic
ov vp
1 yj Mt = p- Pe avas 123
vet Digs =1—— t vA. (123)2
MATHEMATICAL THEORY OF VISCOUS INCOMPRESSIBLE
FLOW
In this chapter we shall be concerned with the theory of the incompressible
Navier-Stokes system (1.2.1), (1.23):
(2.0.1)
ou
Seg
considered in the space-time cylinder Qr = 91 x (0,7), where T > 0 and 2c
IR®(N = 2,3) isa bounded domain occupied by the fluid. We denote the velocity
by w and for simplicity we set p == p/p (it is called kinematic pressure). System
(2.0.1) is equipped with the initial condition
a =f-Vp+vdu,
u(r.0)=u%(x), 2 EM, (2.0.2)
and boundary conditions. We assume that on the whole boundary 09 x (0,7),
the velocity is prescribed:
Ulanxio.7) = (203)
In the case of stationary flow, when O/0t = 0.1 we
get the system
= ula),p = ple), €
in, (204)
-Vpt+vAu, ind
with the boundary condition
lon = 2 (205)
2.1. Function spaces and auxiliary results
In the whole chapter we shall assume that NV =2 or N = 3 and 9c RY isa
bounded domain with a Lipschita-continuous boundary A®2. We shall work with
spaces of continuous and continuously differentiable functions C(@), C¥(M), Cz°(%),
Lebesgue spaces L°(9), Sobolev spaces H1(9), H4(Q), WE*(Q), HQ) =
W*2(Q) and spaces L2(20), H4(22) = W!-42(80) of functions defined on
anFUNCTION SPACES AND AUXILIARY RESULTS
Let us recall that
Hj(O) = {ue HD); ujaQ =O} (2.1.6)
12(Q) and 11°(Q) are Hilbert spaces with the sealar produ
ceshaey= fw hy
and
oss)ayay = f (ov grad w- grad vd, 218)
(ela = foo grad we gon 0) es)
respectively. In virtue of the Friedrichs inequality
a
Uullecay < erllellageay = OF (L radu de) we HR), (2.19)
besides the norm ||- jrijn) induced in the space (9) by the scalar product
Go-Jus(a) also the norm |-Ijygcq) determined by the scalar product
(u geo = ff grad u- grad vde (2.10)
can be used. These two norms are equivalent in the space #73(2).
In this chapter we shall work with \V-component vector-valued functions,
‘whose components are elements of some of the above spaces. We shall use the
following notation:
OF (O) = [DDN = (u= (tay... sun); wi € CFD), F=1,..., N21.)
LQ) = [L7(0)|* (ua,..-, un); ms € £4(2), N},
HQ) = (HQ), HQ) = GA), ete.
In these spaces, the sum of elements and their product with a real number are
defined in an obvious way:
(usu) + (or, uy ey), (24.12)
Murs. stty) = us,
The scalar product in L*(®) is introduced by
(Hur = Dwi erm (2.1.13)6 MATHEMATICAL THEORY OF VISCOUS INCOMPRESSIBLE FLOW
1s -stw) LP),
w= (ues), 8
In a similar way the scalar product ean be defined in H(9). In H}(Q), two
scalar products
(uv }nie = [ DS srad ui+ grad y.de = (24.14)
Sau; Avi
- Leos
ow
(u, v)an(a) = (4, v)L (ny + (MO) (ay, (2.1.15)
induce in H}(®) equivalent norms
lla = (.Wdacays (211.16)
Uelanpa) = (efor
1 means that there exists a constant ¢ > 0 such that
Uellars(e) S llellaecay < ellellgen) Ya € HYCO). (2.117)
In what follows it will be convenient to work with the norm ||-Ifaya) and the
scalar product (yn We shall use the following simplified notation:
(0) = (u2)z2@. (21.18)
(4,0) = (4 e)e2¢0)
(4,2) = (4 2) c0y
ral = Heluscays fall = lege
In order to avoid a misunderstanding in some eases, a scalar product or a norm
will be equipped with a subscript denoting the space considered
Further, let us recall that H4 (82) is the subspace of (82) formed by the
traces on 92 of all functions from H"(Q). Similarly we denote by 3 (8) the
subspace of L?(21) formed by the traces on 89 of all vector functions which
are elements of H!(Q), We use the notation w]0 = (1|89,..., uy 8M) for the
trace of a vector function u = (u;,.-.,y) € H"(®) on 0
By X* we denote the dual of a normed linear space X. The symbol (:)
denotes the duality: if f € X*, g € X, then (fg) = Sle).FUNCTION SPACES AND AUXILIARY RESULTS 7
For q € 17(Q) the symbol dy/@zy (i = 1,...,N) will denote the derivative
in the sense of distributions, i.e., the continuous linear functional on C#(2)
fined by
(She) =~ [ade te= 22) vee cm) (21.19)
Ifve H1(O), then
a
gate (Bea g
For q € 17(9) we understand by grad q the vector whose components are the
distributions 0/Oxj, j= 1,.,N. Let w= (in,.-.,uy) HM). Then
a w= 9 € 240),
Ifu€ HO), then we put
au
au=
where the second order derivatives are considered in the sense of distributions,
For w€ H'(2) we denote by Au = (Au... uy) the vector formed by the
distributions Au,
In the theory of Navier-Stokes problems we shall work with spaces of solenoidal
functions satisfying the condition div = 0. We put
Y= {ue OFM); divu = 0} (21.20)
and denote by V the closure of V in the space H3(2):
vay (21.21)
The space V is clearly a Hilbert one with the sealar produet ((,).
Further, we put
14a) = 24(9y/R! = {ae 149); fade = 2
1410) = Baye’ = {ve (a fade o} (21.22)
Lemma 2.1 Let ¢ € (Hj(®))* and (¢,v) = 0 Vv € V. Then there exists a
function p € L2(Q) such that
(6,0) = (p, div v= [ paivede vee HQ).
In
I means that
(Gv) = —(arad p,v) Vv e CF (O).
and € = -grad p in the sense of distributions.