+ All Categories
Home > Documents > Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In...

Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In...

Date post: 20-May-2020
Category:
Upload: others
View: 14 times
Download: 0 times
Share this document with a friend
53
Comp. Math. & Applications Jingmei Qiu Outline HCL Summary Topic Course on Numerical Methods in Computational Fluid Dynamics Lecture 2: Hyperbolic Conservation Laws Jingmei Qiu Department of Mathematical Science University of Delaware 1 / 53
Transcript
Page 1: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Topic Course on Numerical Methods inComputational Fluid Dynamics

Lecture 2: Hyperbolic Conservation Laws

Jingmei Qiu

Department of Mathematical ScienceUniversity of Delaware

1 / 53

Page 2: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Outline

1 Introduction: definition and examples

2 Development of shocks, weak solutions and the entropysolution.

3 Mathematical properties.

2 / 53

Page 3: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Hyperbolic Conservation Laws

1. Definition and Examples

3 / 53

Page 4: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Hyperbolic conservation laws

Integral form of conservation laws:

d

dt

∫V

udx +

∫∂V

f(u) · nds = 0,

on any control volume V .

• The integral value of u changes in time only due to thenet effect of the flux across the control volume boundary.

• If periodic or compact boundary condition:

d

dt

∫V

udx = 0.

4 / 53

Page 5: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Hyperbolic conservation laws

Differential form of conservation laws:

ut +∇x · f(u) = 0, on ΩI .C . : u(x, t = 0) = u0(x), on Ω.

(1)

• u(x, t): conserved quantities

Rd × R+ → Rm,

• d : the dimension of the problem;• m: the number of components in ~u.

• f(u): flux functions

Rm → Rm

5 / 53

Page 6: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Hyperbolic: 1D system

The 1-D system (1) is hyperbolic if the m×m Jacobian matrix

f ′(u) =∂f

∂u=

∂f1∂u1

· · · ∂f1∂um

· · ·∂fm∂u1

· · · ∂fm∂um

of the flux function has the following property: For each valueof u,

• the eigenvalues of f ′(u) are real,

• the matrix is diagonalizable, i.e. there is a complete set ofm linearly independent eigenvectors.

What if not real? (well-posedness) What if not diagonizable?(weakly hyperbolic)

6 / 53

Page 7: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Example: d = 1, m = 1

Scalar 1D traffic flow model

∂tρ+ ∂x f (ρ) = 0,• ρ = ρ(t, x): car density.

• v(ρ): average velocity ofthe traffic,e.g.v(ρ) = vmax(ρmax − ρ).

• f (ρ) = ρv(ρ)

• Jacobina f ′(ρ) ∈ R.

7 / 53

Page 8: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Example: d = 1, m = 21D shallow water system for fluid in a channel:conservation of

mass : ht + (hu)x = 0,momentum : (hu)t + (hu2 + 1

2gh2)x = 0.

(2)

• h: water height.

• u: horizontal velocity.

• 12gh

2: hydrostaticpressure from verticalintegration.

• How the equations arederived?

• What is f ′(u)? Check thesystem is hyperbolic.

8 / 53

Page 9: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

1D Euler system 1

1I do like CFD. Equation of state for ideal gas law: specific total energy ρE = p

γ−1+ 1

2ρu2; specific

total enthalpy ρH = ρE + p; sound speed c =√∂p∂ρ

.

9 / 53

Page 10: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Hyperbolic: high-D system

The 2-D system in the form of

ut + f(u)x + g(u)y = 0 (3)

with u: R2 × R+ → Rm, f, g : Rm → Rm is hyperbolic if anylinear combination of m×m matrix αf ′(u) + βg′(u), ∀α, β ∈ Rof the flux Jacobians has the following property: For each valueof u,

• the eigenvalues of f ′(u) are real,

• the matrix is diagonalizable, i.e. there is a complete set ofm linearly independent eigenvectors.

10 / 53

Page 11: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Example: d = 2/3, m = 5

2/3D Euler system for air flow, formulated base onconservation of mass (ρ), momentum (ρu) and energy (E ).

ρt +∇ · (ρu) = 0,

(ρu)t +∇ · (ρu⊗ u) +∇p = 0,

Et +∇ · ((E + pI )u) = 0.

u = (ρ, ρu,E )T

f = (ρu, ρu⊗ u + pI,Eu + pu)

11 / 53

Page 12: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

2D Euler system 2

2I do like CFD12 / 53

Page 13: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

13 / 53

Page 14: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

14 / 53

Page 15: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

15 / 53

Page 16: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Solutions of Hyperbolic Conservation Laws

2. Development of shocks, weak solutions and entropy solution.

16 / 53

Page 17: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

1D scalar hyperbolic conservation law

ρt + f (ρ)x = 0.

• Characteristic line in the x − t plane,

dx

dt= f ′(ρ).

Along characteristic lines:

d

dtρ(x , t) = ρt + ρx

dx

dt= ρt + f ′(ρ)ρx = 0

Hence ρ(x , t) remains constant along characteristics.

17 / 53

Page 18: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Illustration in the x − t plane

x

t

dxdt = f ′(ρ)

(0, 0)

18 / 53

Page 19: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Linear advection equationf (ρ) = ρ.

• characteristics: dxdt = 1

• linear advection of initial data with speed 1.

19 / 53

Page 20: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Nonlinear Burgers’ equationf (ρ) = ρ2/2.

• characteristics: dxdt = ρ(x , t).

• depending on the sign of initial data, characteristics go todifferent directions with different speeds.

• when characteristics run into each other: development ofdiscontinuities even from smooth initial data.

20 / 53

Page 21: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Shock developmentExample: Burgers’ equation with sin(x) initial condition, whenthe shock will be developed?

21 / 53

Page 22: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Homework

Use the method of characteristics.

22 / 53

Page 23: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Weak solutions

Solutions in classical sense could fail to exist: development ofdiscontinuities or shocks in future solutions, even with smoothinitial data.

Definition: Weak solution

The function u(x , t) is called a weak solution of the con-servation laws if∫ ∞

0

∫ ∞−∞

(φtu+φx f (u))dxdt+

∫ ∞−∞

φ(x , 0)u(x , 0)dx = 0.

(4)holds for all test function φ ∈ C 1

0 (R×R+), where C 10 is

the space of function that are continuously differentiablewith compact support.

23 / 53

Page 24: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Mathematically the definition of weak solution in the integralform above is equivalent to the solution for the integral form ofthe equation over any choices of spatial and time interval.

• Finite volume scheme: formulated base on the integralform.

• Discontinuous Galerkin method: formulated base onthe weak form (test functions etc.).

24 / 53

Page 25: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Weak solution: R-H jumpcondition

Proposition: Rankine-Hugoniot condition

A function u(x , t) is piecewise smooth and satisfies thePDE strong whenever u ∈ C 1. If the function satisfies theRankine-Hugoniot jump condition along the discontinuitycurve,

s.

= x ′(t) =[|f (u)|]

[|u|], (5)

then u(x , t) is a weak solution of the nonlinear hyperbolicequation. s

.= x ′(t) is the shock speed.

25 / 53

Page 26: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Illustration in the x − t plane

x

tShock

a[

b]

u− u+

x− x+

(0, 0)

26 / 53

Page 27: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Proof base on the integral form

27 / 53

Page 28: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Riemann Problem

• A Riemann problem consists of a conservation lawtogether with piecewise constant initial data having asingle discontinuity.

• The exact solution, involving shocks and rarefactionwaves, to nonlinear Euler equation with Riemann initialdata, can be analytically derived.

• In numerical analysis, Riemann problems appear in anatural way in finite volume methods for the solution ofequation of conservation laws. It is widely used incomputational fluid dynamics and in MHD simulations. Inthese fields Riemann problems are calculated using(approximate) Riemann solvers.

28 / 53

Page 29: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Riemann problem: Example 1

Consider Burgers’ equation, ut + (u2

2 )x = 0 with the initialcondition

u0(x) =

ul = 1, x < 0ur = −1, x > 0

(6)

• Drawing characteristics: colliding into each other(formation of shocks).

• By the R-H jump condition, u(x , t) = u0(x) is the weaksolution of the Burgers’ equation.

29 / 53

Page 30: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Illustration in the x − t plane

x

t

dxdt = 1 dx

dt = −1

(0, 0)

Shock

30 / 53

Page 31: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Riemann problem: Example 2 andnon-uniqueness of weak solution

Consider Burgers’ equation with the initial condition

u0(x) =

ul = −1, x < 0ur = 1, x > 0

(7)

• Drawing characteristics: spreading out.

• A weak solution is the rarefaction wave

u(x , t) =

−1, x < −txt , −t < x < t1, x > t

(8)

• How to derive rarefaction wave solution in the middle?

31 / 53

Page 32: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Illustration in the x − t plane

x

t

Rarefaction

dxdt = −1 dx

dt = 1

(0, 0)

32 / 53

Page 33: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

On the other hand,

• By R-H jump condition, u(x , t) = u0(x) is also a weaksolution of the Burgers’ equation.

• In fact, there are infinitely many weak solutions for thisproblem.

Uniqueness of the weak solution is lost.

33 / 53

Page 34: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Homework

Use the method of characteristics and R-H condition.

34 / 53

Page 35: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Entropy solution

The uniquely exist physically relevant solution among weaksolutions.

• How to define entropies solutions? (vanishing viscositymethod, entropy inequalities)

• What are criteria for selecting entropy solutions amongweak solutions? (entropy conditions)

• What are appropriate spaces (norms) for entropysolutions? (BV norm, L1 and L∞)

35 / 53

Page 36: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Entropy solution

In physical models, the balance laws only come with somephysical viscosity. Conservation laws with viscous terms providemore physically relevant models. For example, in the trafficflow modeling, the ”viscosity” takes the form of slow responseof drivers and automobiles; in the fluid dynamics, the viscositycorresponds to the informal notion of ”thickness”. Forexample, honey has a higher viscosity than water.

36 / 53

Page 37: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Entropy solution: vanishingviscosity method

Definition: Vanishing viscosity principle

Consider the viscous equation

uεt + f (uε)x = εuεxx (9)

An entropy solution of the nonlinear hyperbolic equationis the limit (a.e.) of uε of equation (9) when ε→ 0.

• uε exists for the viscous equation (9).

• The limit of uε exists, as ε goes to zero.

37 / 53

Page 38: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Entropy solution: entropyinequality

Definition: Entropy solution

A weak solution of the nonlinear scalar conservation lawsut + f (u)x = 0 is an entropy solution if for all convexentropy function U(u) with U ′′(u) ≥ 0 and the associatedentropy flux function F (u) with F ′(u) = U ′(u)f ′(u), wehave

U(u)t + F (u)x ≤ 0, (10)

in the distribution sense.

• That is for all φ ≥ 0, φ ∈ C 10 (R× R+), we have

−∫∞

0

∫∞−∞(φtU(u) + φxF (u))dxdt

−∫∞−∞ φ(x , 0)U(u(x , 0))dx ≤ 0.

38 / 53

Page 39: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

39 / 53

Page 40: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

R-H inequality across shock

In particular, across discontinuities in the x − t plane, we haveRankin-Hugoniot like entropy inequality,

−s[|U|] + [|F (U)|] ≤ 0,

for any entropy-entropy flux pairs.

Recall that R-H jump condition across shock is

−s[|u|] + [|f (u)|] = 0.

40 / 53

Page 41: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Example.

Consider Burgers’ equation with Riemann initial data.— Which one is the entropy solution? Rarefaction wave orshock?

41 / 53

Page 42: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Oleinik entropy condition

A discontinuity propagating with speed s = [|f (u)|][|u|] given by the

Rankine-Hugoniot jump condition satisfies the Oleinik entropycondition if for all u between ul and ur ,

f (u)− f (ul)

u − ul≥ s ≥ f (u)− f (ur )

u − ur(11)

where ul and ur are left and right state along the discontinuityrespectively.

Example. Rarefaction wave vs. shock for Riemann problem:which one is the entropy solution?

42 / 53

Page 43: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Lax entropy condition

A discontinuity propagating with speed s given by theRankine-Hugoniot jump condition satisfies the Lax entropycondition if

f ′(ul) > s > f ′(ur ), (12)

where ul and ur are left and right state along the discontinuityrespectively.

Example. Rarefaction wave vs. shock for Riemann problem:which one is the entropy solution?

43 / 53

Page 44: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Entropy Solution ofHyperbolic Conservation Laws

3. Mathematical properties:well-posedness in BV, L1 and L∞ norms.

44 / 53

Page 45: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Bounded Variation (BV) norm

• A function is of bounded variation on a given interval[a, b], if

V ba (f )

.= sup

∑i

|f (xi )− f (xi+1)| <∞,

where xii is any partition of [a, b].

• If u is continuously differentiable, then

V ba (u)

.=

∫ b

a|u′(x)|dx .

• BV norm is an important concept for numerical schemesfor nonlinear equations: TVD, TVB, ENO, WENOschemes.

• Numerical oscillations will contribute to extra functionvariations that are not physical.

45 / 53

Page 46: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Well-posedness in BoundedVariation (BV) norm

Proposition. If u0(x) is a function of locally bounded variationon (−∞,∞), then for each t > 0, u(·, t) is also a function oflocally bounded variation on (−∞,∞), and

TV R−Ru(·, t) ≤ TV R+st

−R−stu0(·),

where s = maxx |f ′(u)|.

46 / 53

Page 47: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

L1 contraction property

Proposition. If u(x , t) and v(x , t) are solutions of the scalarhyperbolic equation with initial data u0(x) and v0(x)respectively, then

‖u(·, t)− v(·, t)‖L1 ≤ ‖u0(·)− v0(·)‖L1 .

Specifically, consider v ≡ 0, then ‖u(·, t)‖L1 ≤ ‖u0(·)‖L1 .

47 / 53

Page 48: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

48 / 53

Page 49: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

49 / 53

Page 50: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

L∞ maximum principle

Proposition. If u(x , t) is a solution of the scalar hyperbolicequation with initial condition u0(x), then

maxx

u(x , t) ≤ maxx

u0(x), minx

u(x , t) ≥ minx

u0(x),

50 / 53

Page 51: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

1D system and high D problems5

• 1D hyperbolic system:— Riemann problem: Hugoniot locus and integral curves.— well-posedness results available only for special class ofinitial data, i.e. initial data with small enough totalvariation. (counter example exists when total variation ofthe solution grows geometrically.3)

• high-D scalar and systems. 4

3https://www.math.psu.edu/bressan/PSPDF/claw-questions.pdf4Measure-Theoretic Analysis and Nonlinear Conservation Laws, Chen,

Torres and Ziemer, 20075https://www.math.psu.edu/bressan/PSPDF/claw-questions.pdf

51 / 53

Page 52: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

Summary

Summary

• Characteristics method: before shocks

• Weak solutions and Rankin-Hugoniot jump condition.

• Riemann problems.

• Entropy solution and entropy inequality

• Well-posedness in bounded variation norm, L1 norm andL∞ norm.

52 / 53

Page 53: Topic Course on Numerical Methods in Computational Fluid ...jingqiu/math817_2019/L2.pdf · In numerical analysis, Riemann problems appear in a natural way in nite volume methods for

Comp. Math.&

Applications

Jingmei Qiu

Outline

HCL

SummaryAcknowledgement:

• Lecture notes by Chi-Wang Shu, when I was a graduatestudent at Brown University.

• Special thanks to Ms. Mingchang Ding (Ph.D. student atUniversity of Delaware) for her help in preparing theseslides.

53 / 53


Recommended