weighted eno schemes - stony brook universitychenx/notes/weighted_eno.pdf · weighted eno schemes...

14
Background WENO Schemes Modified Weights Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics Stony Brook University, The State University of New York February 7, 2014 Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Upload: others

Post on 15-Mar-2021

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Weighted ENO Schemes

Xiaolei ChenAdvisor: Prof. Xiaolin Li

Department of Applied Mathematics and StatisticsStony Brook University, The State University of New York

February 7, 2014

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 2: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

1 Background

2 WENO SchemesBasic IdeaChoice of the WeightsFlux Splitting

3 Modified WeightsMapped WENO SchemesWENO-Z Scheme

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 3: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Background

1D Scalar Hyperbolic Equation:

ut + (f(u))x = 0

Assume the grid is uniform and solve the hyperbolic equation directlyusing a conservative approximation to the spatial derivative.

dui(t)

dt= − 1

∆x(f̂i+ 1

2− f̂i− 1

2),

where ui(t) is the numerical approximation to the point value u(xi, t),and f̂i+ 1

2is called numerical flux.

Question: How to approximate the numerical flux f̂i+ 12

?

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 4: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Basic Idea

xi xi+1 xi+2xi-1xi-2 xi+1ê2xi-1ê2

vi+1ê2-vi-1ê2

+

S

S0

S 1

S 2

Figure: Stencil of fifth order WENO scheme

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 5: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Basic Idea

At smooth region, each sub-stencil gives third (k-th) order numericalfluxes v−

i+ 12

and v+i− 1

2

.

For example, on stencil S1,

v(1)−i+ 1

2

= −1

6v̄i−1 +

5

6v̄i +

1

3v̄i+1

v(1)+

i− 12

=1

3v̄i−1 +

5

6v̄i −

1

6v̄i+1

Similarly, we have

v(0)−i+ 1

2

and v(0)+

i− 12

on stencil S0

v(2)−i+ 1

2

and v(2)+

i− 12

on stencil S2

*The coefficients come from reconstruction process.

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 6: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Basic Idea

Apply a weight to each stencil and a fifth ((2k-1)-th) order WENOscheme is obtained. Assume the weights are w0, w1, w2. Then, werequire

wr ≥ 0,

k−1∑s=0

ws = 1,

for stability and consistency.The fifth order fluxes are given by

v−i+ 1

2

=

k−1∑r=0

wrv(r)−i+ 1

2

, v+i− 1

2

=

k−1∑r=0

wrv(r)+

i− 12

Question: How to choose the weights?

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 7: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Choice of the Weights

If the function v(x) is smooth in all of the candidate stencils, there areconstants dr such that

v−i+ 1

2

=

k−1∑r=0

drv(r)−i+ 1

2

= v(xi+ 12) +O(∆x2k−1),

v+i− 1

2

=

k−1∑r=0

d̃rv(r)+

i− 12

= v(xi− 12) +O(∆x2k−1).

For example, when k=3,

d0 =3

10, d1 =

3

5, d2 =

1

10,

d̃0 =1

10, d̃1 =

3

5, d̃2 =

3

10.

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 8: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Choice of the Weights

In this smooth case, we would like to have

wr = dr +O(∆xk−1).

Form of the Weights:

wr =αr∑k−1s=0 αs

, r = 0, · · · , k − 1

αr =dr

(ε+ βr)2

where βr are the so-called ”smooth indicators” of the stencil Sr.We require

if v(x) is smooth in the stencil Sr, then βr = O(∆x2)

if v(x) has a discontinuity inside the stencil Sr, then βr = O(1)

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 9: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Smooth Indicators

Let the reconstruction polynomial on the stencil Sr be denoted by pr(x).Then, define

βr =

k−1∑l=1

∫ xi+1

2

xi− 1

2

∆x2l−1(dlpr(x)

dxl)2dx

When k=3,

β0 =13

12(v̄i − 2v̄i+1 + v̄i+2)

2 +1

4(3v̄i − 4v̄i+1 + v̄i+2)

2,

β1 =13

12(v̄i−1 − 2v̄i + v̄i+1)

2 +1

4(v̄i−1 − v̄i+1)

2,

β2 =13

12(v̄i−2 − 2v̄i−1 + v̄i)

2 +1

4(v̄i−2 − 4v̄i−1 + 3v̄i)

2

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 10: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Basic IdeaChoice of the WeightsFlux Splitting

Flux Splitting

Lax-Friedrichs Splitting:

f±(u) =1

2(f(u)± αu),

where α = maxu|f ′(u)| over the relevant range of u.FD WENO Procedure with Flux Splitting:

Identify v̄i = f+(ui) and use WENO procedure to obtain v−i+ 1

2

, and

take f̂+i+ 1

2

= v−i+ 1

2

Identify v̄i = f−(ui) and use WENO procedure to obtain v+i+ 1

2

, and

take f̂−i+ 1

2

= v+i+ 1

2

Form the numerical flux as f̂i+ 12

= f̂+i+ 1

2

+ f̂−i+ 1

2

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 11: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Mapped WENO SchemesWENO-Z Scheme

Mapped WENO Schemes

gr(w) =w(dr + d2r − 3drw + w2)

d2r + w(1− 2dr)

α∗r = gr(wJSr )

wMr =

α∗r∑2s=0 α

∗s

Advantage: fifth order accuracy at critical pointsDisadvantage: the weight of the stencil that contains discontinuitybecomes larger after the map; more computational cost.

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 12: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Mapped WENO SchemesWENO-Z Scheme

WENO-Z Scheme

τ5 = |β0 − β2|

αzr = dr(1 + (

τ5βk + ε

)q)

wzr =

αzr∑2

s=0 αzs

Advantage: fifth order accuracy at critical points with q=2;computationally more efficient than mapped WENO.

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 13: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Mapped WENO SchemesWENO-Z Scheme

Bibliography I

[1] B. Cokcburn, C. Johnson, C.-W. Shu, and E. Tadmor, Advancednumerical approximation of nonlinear hyperbolic equations, Ed. A.Quarteroni, Lecture Notes in Mathematics, vol. 1697, Springer,1998.

[2] C.-W. Shu, High order weighted essentially non-oscillatoryschemes for convection-dominated problems, SIAM Review, v51(2009), pp.82-126.

[3] A. K. Henrick, T. D. Aslam, J. M. Powers, Mapped weightedessentially non-oscillatory scheme: Achieving optimal order nearcritical points, Journal of Computational Physics 207 (2005),pp.542-567.

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes

Page 14: Weighted ENO Schemes - Stony Brook Universitychenx/notes/weighted_eno.pdf · Weighted ENO Schemes Xiaolei Chen Advisor: Prof. Xiaolin Li Department of Applied Mathematics and Statistics

BackgroundWENO Schemes

Modified Weights

Mapped WENO SchemesWENO-Z Scheme

Bibliography II

[4] R. Borges, M. Carmona, B. Costa, W. S. Don, An improvedweighted essentially non-oscillatory scheme for hyperbolicconservation laws, Journal of Computational Physics, 227, (2008)pp.3191-3211.

Xiaolei Chen Advisor: Prof. Xiaolin Li Weighted ENO Schemes