TIME INTEGRATION

S. Roller, [email protected], Raphael Haupt, M.Sc. ([email protected])

Institute of Software Methods for Product Virtualization (SP), DLR
 Modelling

Reality

Physical Model

Mathematical Model

Numerical Approximation

Simulation code

Usage of the simulation code

Analysis of results

2 HPCFD01 - CFD - An Overview

 Physical Continuum Model

§ Newtonian mechanics

§ Conservation laws:
§ Conservation of mass
§ Conservation of momentum
§ Conservation of energy

3 HPCFD01 - CFD - An Overview

 Conservation Laws

§ Quantity can neither vanish nor appear @A

<latexit sha1_base64="EMmrtiGNirWcigpvKP9xU+gRZ7c=">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</latexit>

~n
<latexit sha1_base64="Eq7Ci93vl5JRywtTO6WYLfGClQo=">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</latexit>
<latexit

§ Observing a given volume:

§ Quantity only changed by transport through the surface A
<latexit sha1_base64="PgW2ZYpeAi+iQXQqE7VSXvYchk4=">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</latexit>
<latexit

§ Mathematical:

Z I
@
u(~x, t)dA = f~(u)~nds
<latexit sha1_base64="9NLo+X4CSU5yNwb9HS+/r2YijeA=">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</latexit>
@t A @A

4 HPCFD01 - CFD - An Overview

 Conservation of Mass

§ Conserved: Mass density ⇢

<latexit sha1_base64="2VVbe9OvnocbBJaY+KdA4G5dZmg=">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</latexit>

§ Changed by:
§ Convection f~ = ⇢~v
<latexit sha1_base64="pQ85aWSlg2t7+I0nm6EnhkGpIZ0=">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</latexit>

Z I
@
§ Integral form: ⇢(~x, t)dA = ⇢~v · ~nds
<latexit sha1_base64="ZfF3d8RxMkioXr94/t0BvEej6Ls=">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</latexit>
@t A @A

@⇢(~x, t)
§ Differential form: + r · (⇢~v ) = 0
<latexit sha1_base64="oxXpGUn0yjzKWV+C1UkZ2EgcPxs=">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</latexit>
@t
5 HPCFD01 - CFD - An Overview
 Conservation of Momentum

§ Integral form:
pres. fri.
Z I ⇣
convect.
⌘ Z
@
m(~
~ x, t)dA = m
~ ~v + pI ⌧ ~nds + F~ e dA
<latexit sha1_base64="ZPbYC2WH0+ZFFx6rXUL3Q93ideg=">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</latexit>
@t A @A A
§ Differential form:

convect. pres. fric.

@ m(~
~ x, t)
~ ~v + pI) = r · ⌧ + F~ e
+ r · (m
<latexit sha1_base64="8egCZ387UNG/pQHMxeeSa6yD4E4=">AAADcXichVFdaxQxFL2z40dbq13tk/gSughbKstMERRaofhR9EGo4LaFppZMNruGnS8y2aV1mL/jf9K/4asPnsSppVZphsy9Offcc29ykzLVlY2ib0EnvHHz1u2FxaU7y3fvrXTvP9ivipmRaiiLtDCHiahUqnM1tNqm6rA0SmRJqg6S6SsXP5grU+ki/2jPSnWciUmux1oKC+ik+5WPjZA1L4WxWqSMz5Wss6bv7WnzhNn15iJqG7bBeC6SVDAuR4Vl/TaBS22k9+eOUzJeoKzrqr7w3jXNOntxWeBfPG7FrPGlnOBu80mddHvRIPKLXXXi1ulRu/aK7nfiNKKCJM0oI0U5WfgpCarwHVFMEZXAjqkGZuBpH1fU0BJyZ2ApMATQKf4TnI5aNMfZaVY+W6JKim2Qyegx9q5XTMB2VRX8CvYn9hePTf5bofbKrsMz2ASKi17xPXBLn8G4LjNrmee9XJ/pbmVpTM/9bTT6Kz3i7in/6LxGxACb+gijN545gUbiz3O8QA47RAfulc8VmL/xCFZ4q7xK3ioK6BlY9/roB2OO/x7qVWd/cxBHg/jD097Oy3bgC/SI1qiPqT6jHXpLe+hDBsvBZrAVbHd+hA9DFq79pnaCNmeVLq1w4xeARMvR</latexit>
@t

6 HPCFD01 - CFD - An Overview

 Conservation of Energy

§ Integral form:

Z I conv pres. fric. heat Z

@ .
e(~x, t)dA = (e · ~v + p · ~v ⌧~v rT )~nds + (F~ e · ~v + Q)dA
<latexit sha1_base64="kX7/hO/F7/LJ+WZ4mArj9oKMCsU=">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</latexit>
@t A @A A

§ Differential form:
@e(~x, t)
+ r(e + p)~v = r · ⌧~v + r · (rT ) + F~ e · ~v + Q
<latexit sha1_base64="kCx+ycdvWqnGqE4PXSbZr8seTpI=">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</latexit>
@t

7 HPCFD01 - CFD - An Overview

 Equation of State

§ Relation of thermodynamic quantities density, pressure and temperature

§ Simplest thermodynamic model is the ideal gas assumption:

p
=R·T
⇢
gas constant
<latexit sha1_base64="dQj/6lPy/oWqEB7XkJmP5lEiHSU=">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</latexit>

8 HPCFD01 - CFD - An Overview

 Compressible Navier-Stokes Equations

@⇢(~x, t)
+ r · (⇢~v ) = 0
<latexit sha1_base64="I/v/rYzbRvQcBiWjR59nNn3x5MQ=">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</latexit>
@t
@ m(~
~ x, t)
+ r · (m~ ~v + pI) = r · ⌧
<latexit sha1_base64="Ls/XFY5rmRVMBqEU0IioS4xjM8U=">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</latexit>
@t
@e(~x, t)
+ r · (e + p)~v = r · ⌧~v + r · (rT )
@t
<latexit sha1_base64="NAVN81d/WWE6cw4xo55oS7Z/cE8=">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</latexit>

+Equation of state ⇣ ⇢ 2⌘
- Perfect gas: p=( 1) · e |~v |
<latexit sha1_base64="862CZ2La2n//u0GmF+mfhJ8AU6U=">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</latexit>
2
9 HPCFD01 - CFD - An Overview Image: game-icons.net, Lorc
 Simplification Overview

Compressible Navier-Stokes equations

include friction and thermal conduction

hyperbolic – parabolic

Re ! 1
<latexit sha1_base64="I5luM6vLngbZrl954Tr/g73LDLE=">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</latexit>
Ma ! 0
<latexit sha1_base64="HqS3/c/EXuJETRyDySrCTRr3Ki8=">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</latexit>

Incompressible Navier-
Euler equations
Stokes
Gas dynamics
equations
hyperbolic
parabolic - elliptic

10 HPCFD01 - CFD - An Overview

 Gas Dynamics: The Euler Equations

§ Hyperbolic system (wave transport)

@⇢(~x, t)
+ r · (⇢~v ) = 0
<latexit sha1_base64="rIqKqDdB5/3K6TvRh3MndGQBJak=">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</latexit>
@t
@ m(~
~ x, t)
+ r · (m~ ~v + pI) = 0
<latexit sha1_base64="oQO8+ji2GV7U1LvvppEyUgrjk+E=">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</latexit>
@t
@e(~x, t)
+ r(e + p)~v = 0
@t
<latexit sha1_base64="5+0Kqp636VyWrt6Z1WTVQJXcXjE=">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</latexit>

+Equation of state (usually ideal gas)

11 HPCFD01 - CFD - An Overview Image: game-icons.net, Lorc

 Define state u, flux f

State Flux
2 3 2 3
⇢~v
<latexit sha1_base64="LPIW2pzSFEyfwJFF4P9q9htT4RI=">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</latexit>

⇢(x, t)
~u(x, t) = 4m(x, t)5 f~(~u) = 4m
~ ~v + pI 5
e(x, t)
<latexit sha1_base64="F16TeNiI9lJEoBpVJDLtfVd3ahQ=">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</latexit>
(e + p)~v

12 HPCFD01 - CFD - An Overview

 Compact Notation: 3 ways to say the same

d
X
@~u(~x, t) @ f~i (~u)
+ =0
@t i=1
@xi
<latexit sha1_base64="BEByeRArX6YE2x0yFAvEyOoFYwU=">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</latexit>

~ut + r · f~ = 0
<latexit sha1_base64="p0kFveUOIxADLvjG9S0rfRFQ2BU=">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</latexit>

~ut + f~1 x + f~2 y = 0

<latexit sha1_base64="WSvier6Tm1/Q/YlOyj5ob2YktEM=">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</latexit>

13 HPCFD01 - CFD - An Overview

 Formulating with velocity in component notation

<latexit sha1_base64="bVaCW7wfRI3putCVbO82ylxUikk=">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</latexit>

2 3 2 3
⇢ ⇢
6m1 7 6⇢v1 7
~u = 6 7=6 7
4m2 5 4⇢v2 5
e e
2 3 2 3
<latexit sha1_base64="6jW4A773QmaDF1rxSOna0dWAcw8=">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</latexit> <latexit sha1_base64="MGUnf4BBoWLYAdaVmi03EMUQn50=">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</latexit>

⇢v1 ⇢v2
6 ⇢v12 + p 7 6 ⇢v1 v2 7
f1 (~u) = 6
4 ⇢v1 v2 5
7 f2 (~u) = 6 2
4 ⇢v2 + p 5
7

(e + p)v1 (e + p)v1
14 HPCFD01 - CFD - An Overview
 Discretization

~ut + r · f~ = 0
<latexit sha1_base64="p0kFveUOIxADLvjG9S0rfRFQ2BU=">AAAC8HichVHRShtBFD1uWzVpa9P66MtgEAQhbEqrvhTEaulLIUJjAkkIs5NJumSzu8xOAhryBf2BvhVfffO1fkn7LX3o2elaUCnOMnvvnHvumXvnBmkUZtb3fy55jx4/WV5ZLZWfPnu+9qLy8tVplkyN0k2VRIlpBzLTURjrpg1tpNup0XISRLoVjN/n8dZMmyxM4s/2LNW9iRzF4TBU0hLqV7a6M63m00Xfih3RjWUQya4aJFY4fLgQ74RfLvcrVb/muyXuO/XCqaJYjaTyC10MkEBhigk0Ylj6ESQyfh3U4SMl1sOcmKEXurjGAmXmTsnSZEiiY/5HPHUKNOY518xctuItEbdhpsAW9wenGJCd36rpZ7S/uc8dNvrvDXOnnFd4RhtQseQUPxG3+ELGQ5mTgnlTy8OZeVcWQ+y7bkLWlzok71P90zlixBAbu4jAsWOOqBG484wvENM2WUH+yjcKwnU8oJXOaqcSF4qSeoY2f33WwzHX7w71vnP6ulbfrb09eVM9OCwGvooNbGKbU93DAT6iwToUvuIKP3DtGe+b9927+Ev1loqcddxa3uUfyfycsg==</latexit>

r · f~
<latexit sha1_base64="iRUcP1x8e9fvrjtTbfSjXbbitF8=">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</latexit>

~ut =
<latexit sha1_base64="kIBPr2Xt/G7IDeFuPWTWwbWi+oY=">AAADBnichVFNSxxBEH2OxqhJzKrHXAYXwUBYZiUaQQTxI3gR/Miq4MoyM/aOzc7ODD29i7rs3b/hH8hNcs0tVz2a35KDb9o2JJFgDz1V9arqdX0EWSxz7Xl3A87g0IvhlyOjY69evxl/W5qY3M/TjgpFLUzjVB0Gfi5imYialjoWh5kSfjuIxUHQWiv8B12hcpkmX/R5Jo7bfpTIpgx9TahRWqp3Rdjr9BvaXXZ3N/fqsWjqWQv+YZz1P+i6ktGpfm+F6zZKZa/imeM+VapWKcOe7bT0E3WcIEWIDtoQSKCpx/CR8ztCFR4yYsfoEVPUpPEL9DHG3A6jBCN8oi3+I1pHFk1oF5y5yQ75SsyrmOlihvezYQwYXbwqqOeUv3gvDBb994WeYS4qPKcMyDhqGLeIa5wy4rnMto18rOX5zKIrjSYWTTeS9WUGKfoMf/Os06OItYzHxYaJjMgRGLvLCSSUNVZQTPmRwTUdn1D6RgrDklhGn3yKspg+6+Gaq/8u9amyP1epLlTmdz6WV1btwkfwDtOY5VY/YQWb2GYdIa7wAze4dS6dr8618+0h1BmwOVP46zjf7wES/Kbp</latexit>

~ut = RHS (~u (~x, t))

15 HPCFD01 - CFD - An Overview

 Numerical Quadrature

§ Let us first consider integration of:

du
<latexit sha1_base64="4rSrRHPfJGWVtzNOCoQ6xa1bDzk=">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</latexit>

= RHS(t)
dt
§ This is referred to as quadrature problem, and is solvable by the explicit
integral:
<latexit sha1_base64="e7x4BG917DbGZC5FySw9BRXe/MU=">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</latexit>

Z t
u(t) = u0 + RHS(✓)d✓
t0

16 HPCFD01 - CFD - An Overview

 Numerical Quadrature

§ If the antiderivative of f can not be found, the idea of numerical quadrature is

to approximate it by something that is known.

§ The basic idea is to replace the function on small intervals by a simpler

function that can be integrated. Often polynomials are used for this.

§ In the following, we‘ll have a brief look at some procedures to obtain such an
approximation.

17 HPCFD01 - CFD - An Overview

 Mid-Point Rule

§ Use the value of f at a set of points ti, and build an area approximation by
multiplying ti with a Δt:

Z
<latexit sha1_base64="LIbro/OmUkDyMt29xpO1NpWrUps=">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</latexit>

t N
X
RHS(✓)d✓ ⇡ RHS(ti ) · t
t0 i=1

18 HPCFD01 - CFD - An Overview

 Trapezoidal Rule

§ Replacing the constant function value in the midpoint rule by a linear

approximation of the function yields the trapezoidal rule.

Z
<latexit sha1_base64="puB8RSzQtWUUlp035qZ2ffZYZ6Y=">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</latexit>

t N
X t
RHS(✓)d✓ ⇡ (RHS (ti ) + RHS (ti 1 )) ·
t0 i=1
2

19 HPCFD01 - CFD - An Overview

 Convergence Order

§ Smaller errors are achieved by more steps.

§ The convergence order describes how fast the error diminishes with
decreasing step size.

e = const · hp
e : error
h : stepsize
<latexit sha1_base64="N1SleAJ0tZoINVx6ytAjl61rmYA=">AAACLnicbZBNSwMxEIazflu/qh69BIviqeyKoAiCKIJHBatCt5ZsdtYGs0lIZoVa/EVe/Ct6EFTEqz/DdO3Brzm9ed4ZMvMmRgqHYfgcDA2PjI6NT0xWpqZnZueq8wunTheWQ4Nrqe15whxIoaCBAiWcGwssTyScJVf7ff/sGqwTWp1g10ArZ5dKZIIz9KhdPVgFukO5Vg5pzFONtHNhaBxXPN+mYK225avjXw7BOHEDJTAeaJtCabertbAelkX/imggamRQR+3qY5xqXuSgkEvmXDMKDbZ6zKLgEm4rceHAMH7FLqHppWI5uFavPPeWrniS0sxvlmmFtKTfJ3osd66bJ74zZ9hxv70+/M9rFphttXpCmQJB8a+PskJS1LSfHU2FBY6y6wXjVvhdKe8wyzj6hPshRL9P/itO1+tRWI+ON2q7e4M4JsgSWSZrJCKbZJcckiPSIJzckQfyQl6D++ApeAvev1qHgsHMIvlRwccnSKKkag==</latexit>
p : order

20 HPCFD01 - CFD - An Overview

 High-Order Integration

§ By adding more function values to the interval approximation, we arrive at

smoother and smoother approximations of f.

§ Prescribing the points, where the function is evaluated, equidistantly results

in quadrature rules, that are referred to as Newton-Cotes formulae.
§ The Trapezoidal rule is the one of these

§ For n function values they yield an approximation of at least order n-1.

21 HPCFD01 - CFD - An Overview

 Simpson’s rule

§ For example using a polynomial of degree 2 for the local approximation the
Simpson rule is found

Z N ✓ ✓ ◆ ◆
<latexit sha1_base64="05wLRjFXVqRDpuuMwg4PH3wJ2+o=">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</latexit>

t X ti + ti 1 t
RHS(✓)d✓ ⇡ RHS (ti 1 ) + 4RHS + RHS (ti ) ·
t0 i=1
2 6
22 HPCFD01 - CFD - An Overview
 Chebyshev Quadrature

§ Prescribing weights of summation as uniform instead of evaluation points:

Z
<latexit sha1_base64="mOvsTP7Os+vPOM4h63Hd13LeIpQ=">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</latexit>

1 Xn
2
RHS(✓)d✓ ⇡ RHS(ti )
1 n i=1
§ Results in non-equidistant function evaluation (relatively complicated
computation of point locations)
§ These methods are referred to as Chebyshev quadrature
§ Approximation order is improved over Newton-Cotes: at least an order of n is
achieved with n points

23 HPCFD01 - CFD - An Overview

 Gauss Quadrature

§ By freely choosing the evaluation points and the integration weights,

Gaussian quadrature rules are obtained.

§ They provide an approximation order of 2n – 1 for n points, if the integrated

function is sufficiently smooth.

§ If choice of integration points is free, preferred method for numerical

integration!

24 HPCFD01 - CFD - An Overview

 Gauss Quadrature with 2 Points

§ For two points, we get for example:

Z
<latexit sha1_base64="c5YZXPsAYgvvOoCKU3gS9rxXkdc=">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</latexit>

r r !
t
t t0 t0 + t 1 t t0 + t 1 t
RHS(✓)d✓ = RHS( ) + RHS( + )
t0 2 2 3 2 2 3 2

§ All these quadratures can be viewed as local approximations by polynomials,

however clever selection of evaluation points allows an increased accuracy
in the numerical integration.

25 HPCFD01 - CFD - An Overview

 Variations on the Gaussian Integration

§ Sometimes some points need to be fixed, but most can be chosen freely

§ For these kind of integrations, variants of the Gaussian integration can be

used

§ Reduces the order of accuracy

§ Notable example: Gauss-Lobatto integration (includes end points of the

interval)

26 HPCFD01 - CFD - An Overview

 Romberg’s Method

§ All methods presented so far, offer a fixed accuracy.

§ Romberg‘s Method allows the iterative refinement of the integration rule, until
a certain accuracy is achieved.

§ It uses the trapezoidal rule and keeps adding more and more values, while
reusing those from previous iterations.

t t0
§ The interval length is halved in each iteration: tn =
<latexit sha1_base64="0JYivUovHLEGACrF9YujG5E+33U=">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</latexit>
2n
27 HPCFD01 - CFD - An Overview
 Romberg‘s Method (Details)

§ First step: trapezoidal rule

<latexit sha1_base64="Ac+dTU3Kh1I93dqrWUShCAXatYY=">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</latexit>

R(0, 0) = t1 (RHS(t) + RHS(t0 ))

§ In the refinement steps we first need to evaluate the function at additional
points and reuse the result from the previous iteration
n 1
<latexit sha1_base64="ddAOjIxYxzfKrGBvJ0W8vhnRdWA=">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</latexit>

2X
R(n 1, 0)
R(n, 0) = + tn RHS(t0 + (2k 1) tn )
2
k=1

28 HPCFD01 - CFD - An Overview

 Romberg‘s Method (continued)

§ The obtained results can then be used for polynomial extrapolation without
evaluating the function to integrate:

R(n, m 1) R(n 1, m 1)
R(n, m) = R(n, m 1) +
<latexit sha1_base64="a9KUyK14GjPaBLcF40b1cJjunbs=">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</latexit>
4m 1

§ Of these extrapolations we can do as many, as we have iterations (𝑚 ≤ 𝑛)

§ The convergence order is 2m+2

29 HPCFD01 - CFD - An Overview

 Numerical Methods for More General ODEs

§ Finding the integral curve of an ODE is similar to the identification of the

quadrature.
<latexit sha1_base64="jKFIaw8XK/ZH3L73n5s9aP5bDqs=">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</latexit>

Z t
u(t) = u0 + RHS(✓, u)d✓
t0

§ However, we now have to deal with the additional complication, that f

depends on x itself.

30 HPCFD01 - CFD - An Overview

 Euler (forward) Method

§ The Euler method is a first order approximation and uses the value of f as
slope for a specific step width Δt.

§ That is, starting from the initial condition the next point on the approximated
integral curve is found by:

u(t) = u0 + RHS(t0 , x0 ) · (t
<latexit sha1_base64="+truCbsGYXxvVFD4TjM1/x7QwQQ=">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</latexit>

t0 )

31 HPCFD01 - CFD - An Overview

 Euler Method Stability

§ Besides the low accuracy of the Euler method, it also has a very limited
stability region.

du
<latexit sha1_base64="Z8qi0VFSZ8GpZB+C/AWtaBpXT7M=">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</latexit>

§ When considering the linear ODE = k ·u,

dt
we see, that the Euler method is only stable, if -2 < kΔt < 0, as otherwise the
overshooting by the method will grow ever larger

32 HPCFD01 - CFD - An Overview

 Backward Euler

§ The stability limitation can be overcome by using an implicit formulation of

the Euler method and evaluating the function at the end point instead of the
starting point to define the slope:

u(t) = u0 + RHS(t, u(t)) · (t

<latexit sha1_base64="cbhOTFpg3vwAjPUge2Q1RUeG004=">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</latexit>

t0 )
§ However, this results in the need to solve an equation for x(t), as it appears
on both sides.

33 HPCFD01 - CFD - An Overview

 Implicit Methods

§ Using Backward Euler overcomes the stability issue, but usually the resulting
equation can not be solved analytically.

§ Thus, an iterative approach to the solution of the arising equation is normally

required.

34 HPCFD01 - CFD - An Overview

 Trapezoidal Rule

§ A refinement of the implicit backward Euler method can be achieved by the

trapezoidal rule.

§ For this we consider the averaged values from both interval ends:

<latexit sha1_base64="/0YED406aaW0XGD8gqMb8lAWHvw=">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</latexit>

RHS(t0 , u0 ) + RHS(t, u(t))

u(t) = u0 + · (t t0 )
2

35 HPCFD01 - CFD - An Overview

 Mid-Point Rule

§ A refinement of the Euler method, that remains an explicit approach, is the

Mid-Point rule.
§ Here the slope is approximated by an intermediate point
§ Intermediate value (Euler fwd. for half step only):

t + t0
<latexit sha1_base64="3pfOZj3gsUng74TNRLTTqnl5DW4=">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</latexit>

u(t1/2 ) = u(t0 ) + RHS(t0 , u(t0 ))

2
§ New value (result):

t0 ) · RHS(t1/2 , u(t1/2 ))
<latexit sha1_base64="WemedTmouZQynGLbrg2uFehE2m4=">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</latexit>

u(t) = u(t0 ) + (t

36 HPCFD01 - CFD - An Overview

 Runge-Kutta

§ The principal of the Mid-Point rule to use intermediate evaluations to improve

the accuracy can be generalized to use more points to increase the
approximation order.
§ These kind of methods are known as Runge-Kutta methods.
§ The classical Runge-Kutta Method is the RK4 method using 4 substeps.

37 HPCFD01 - CFD - An Overview

 Classical Runge-Kutta (KR4)

§ The final approximation is computed with the help of four intermediate steps:

t t0
<latexit sha1_base64="Q8enY4SQftP1+JmQuejKOU8bq0U=">AAADEXichVHLbtNAFD01rza8AizZjIiQEiEiO5TCBqniUbFBKhJpKzWVNZ5MgmXHtsaTSiXKV/Ab/AA7xJYdOx6fwoIzUxcJKtRrjc+dc+89c2duUuVpbcPw20pw7vyFi5dW11qXr1y9dr194+ZOXc6N0kNV5qXZS2St87TQQ5vaXO9VRstZkuvdJHvm4ruH2tRpWbyxR5U+mMlpkU5SJS2puL0179qeeCIIcdgT98RoYqRaWHFfkFguNpZipMalFd0sjhgeZPHgGB4Qsni913IWtzthP/QmTjtR43TQ2HbZ/o4RxiihMMcMGgUs/RwSNb99RAhRkTvAgpyhl/q4xhIt1s6ZpZkhyWb8T7nbb9iCe6dZ+2rFU3Iuw0qBu1xbXjFhtjtV06+Jv7jeeW763xMWXtl1eERMqLjmFV+Rt3jLjLMqZ03mSS9nV7pbWUzw2N8mZX+VZ9w91R+d54wYcpmPCLzwmVNqJH5/yBcoiEN24F75REH4G4+J0qP2KkWjKKlniO712Q/HHP071NPOzqAfbfQfvl7vbD5tBr6K27iDLqf6CJt4iW32ofABX/EDP4P3wcfgU/D5ODVYaWpu4S8LvvwGZgOjyw==</latexit>

u(t) = u(t0 ) + · (k1 + 2k2 + 2k3 + k4 )

6
§ With h=t t0
§ The substeps read:

38 HPCFD01 - CFD - An Overview

 Classical Runge-Kutta (RK4)

<latexit sha1_base64="zgWupg/0/pnoDI0VO3+4ZU6mYHs=">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</latexit>

k1 = RHS(t0 , u0 )
h h
k2 = RHS(t0 + , u0 + k1 )
2 2
h h
k3 = RHS(t0 + , u0 + k2 )
2 2
k4 = RHS(t, u0 + hk3 )

39 HPCFD01 - CFD - An Overview

 Butcher Barrier

§ Increasing the convergency order of RK methods by more steps is only

reasonable up to some limit.
§ Due to the observation, known as Butcher barrier, the amount of necessary
intermediate steps grows faster than the convergency order.
§ (For five and more stages, there is no explicit RK that achieves the same
convergency order as it has number of steps).

40 HPCFD01 - CFD - An Overview

 Multistep Methods

§ While the RK methods make use of intermediate evaluations (sub-steps), the

linear multistep methods attempt to improve the accuracy by using
information from previous time steps.
§ Three families of multistep methods are commonly used:
§ Adams-Bashforth
§ Adams-Moulton
§ Backwards Differentiation Formulas

41 HPCFD01 - CFD - An Overview

 Two-Step Adams Bashforth

§ An example for a multistep method is the two-step Adams-Bashforth method,

which requires two time step information to compute the next time step:

3h h
<latexit sha1_base64="poc3CPRPNEpAAAxTTVL3JiKEjJ8=">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</latexit>

ui+1 = ui + · RHS(ti , ui ) · RHS(ti 1 , ui 1 )

2 2

§ The Adams-Bashforth methods can be understood as an interpolation

polynomial through the known time-steps.

42 HPCFD01 - CFD - An Overview

 Variable Order Approximation

§ Similar to the Romberg Method it is also possible to use an adaptive scheme

to find the integral curve of the ODE.

§ Schemes like the Bulirsch-Stoer algorithm rely on a Richardson extrapolation

to refine the approximation until a given error is reached.

43 HPCFD01 - CFD - An Overview