M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne
description
Transcript of M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne
![Page 1: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/1.jpg)
1
M. Darbani, A. Ouahsine, P. Villon
University of Technology of Compiègne Roberval laboratory UMR-CNRS 6253, France
MULTIPHYSICS-200909-11 December 2009, LILLE ,France
Natural Elements Method for ShallowWater Equations
![Page 2: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/2.jpg)
2
Contents
1)1) Problem FormulationProblem Formulation
2)2) Shape FunctionsShape Functions
3)3) Shallow water Equations and Time discretizationShallow water Equations and Time discretization
4)4) Nodal Integration Nodal Integration
5)5) Numerical tests and results application to Dam BreakNumerical tests and results application to Dam Break
Problem Shape Function SWE & Time discretizatio Nodal Integration Numerical test Appllication
![Page 3: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/3.jpg)
3
Dam Break
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Breaking waves
Large deformation
Finite Elements method with meshing is not always convenient.
![Page 4: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/4.jpg)
4
Why meshless method ?
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
In NEM method the possibility of
treating the problems is easier for large
deformations than in the finite elements method
Ability to insert or remove the nodes easier
Example: Domain enrichment
Shape functions depend only on the positions
of the nodes.
![Page 5: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/5.jpg)
5
Voronoi Diagram and Delaunay triangles
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Delaunay Triangulations
and Circumscribed circles
Voronoi diagram
Cloud of the nodes
![Page 6: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/6.jpg)
6
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Mathematical formulation
ijkXXdXXdXXdRxT kjim
ij ),,(),(),(:
jixxdxxdRxT jim
i ),,(),(:
The Natural Neighbors of a node are the nodes associated with neighboring Voronoi cells. They are connected to the node by an edge of Delaunay triangle.
The first order and second order cells of the Voronoi diagram are defined mathematically by :
Cells 1st and 2d order
Natural Neighbor
![Page 7: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/7.jpg)
7
Example
![Page 8: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/8.jpg)
8
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Sibsonian Shape Function
Example
nxi
i x xii 1x
K(x) ,K KK
3Area(cdef )(P)Area(abcd)
![Page 9: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/9.jpg)
9
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Non-Sibsonian Shape Function
1
( ) ( )( ) , ( )( )( )
i i
i ini
jj
x l xx xd xx
di(x): Euclidian distance between points x and node ni
li(x): length of the Voronoi edge associated to x and node ni
![Page 10: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/10.jpg)
10
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Properties of NEM Shape Function
property of the Kronecker delta
Partition of unity
Local co-ordinate propertyLinear consistency
i j ij
1 , i j(x )
0 , i j
n
ii 1
(x) 1
n
i ii 1
x (x)x
![Page 11: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/11.jpg)
11
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Properties of NEM Shape Function
is C1 in every points of Circumscribed circles
is only continuous in every points of Circumscribed circles
sibi (x) :
non sibi (x):
Remark :
![Page 12: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/12.jpg)
12
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Shallow Water Equations
0(u, v) f 2 sin v
),0(),(),(),( TttVt DD xxxv
D Stand for the Dirichlet
portion of the boundary
Governing and Continuty Equations for Incompressible Fluid
0d(hv)f (h ) g h .(h ) in *(0,T)
dt
.(h ) 0 in *(0,T)t
k v v
v
Boundary Conditions
h(x,y,t)=η(x,y,t)+H0(x,y)
H(x,y,t)=h(x,y,t)+hb(x,y)
![Page 13: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/13.jpg)
13
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Time discretization
n n 1
d ( ( . ) ) Material derivativedt t
(x) (X) Lagrangian formulationt
v v v v
v v
dtXd
txd
tXx nnnn
*1
**1 )()()()( vvvvvvv
kk
kknn
kk
kknn
td
tX
td
tx
)()()(
)()()(
*1*
1
**
vvvv
vvvv
n 1
n
X Position of Point at t
x Position of Point at t
n 1
n
*
k
Gauss point at t
Gauss point at t
v Test functionWeight functionof Gauss
![Page 14: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/14.jpg)
14
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Iterative process
n
n
tatPtGauss
tatPtGauss
.
. 1
Determine vertex of triangle at the time step t (n-1) :X(n-1)=x(t)-v(n-1)t
Finding velocity at the point old with interpolation :
V(old )=ΣNi(old )vi
Find New at t (n-1) step withNew=ξ-v (old )
If | old - New |/ | old |<10-5
New = old
New=ξ-v (old )
![Page 15: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/15.jpg)
15
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
The weak form of the Diffusion Term leads to the following integral :
The above integral can be approximated by the following assumption with considering the method SCNI [Chen 2001] :
i
ij i jDif N N d
i
ij app i jΩi σΩ σΩ
1(Dif ) = nN dγ nN dγmes( )
j
With method of Gauss for numerical integration we have:
gi
Nedge Nedge
i i g g i gk 1 k 1 x edgeσΩ edge
nN d nN d n(x )ω N (x )
Diffusion Term
n i: External normal
Xg &ω i :Integration point and weight of Gauss
Mes(ω i): Area of ω i
![Page 16: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/16.jpg)
16
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllicatio
i j i iΩj
N N dΩv =mes(Ω)v
The weak form of the Coriolis Term leads to the following integral:
i j i i j iΩ Ωj j
N N dΩv = N ( N )dΩv
iN d mes( )
Finally, the approximated Coriolis term will become :
By using we obtain :ii
N =1
and by using the second propriety we will obtain i j ijN (x )=δ
Coriolis Term
i j i i ij
N N d v N d v
![Page 17: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/17.jpg)
17
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
The global matricial form is :
n 1 n n n 1
(n 1)
(n)
[M ] : Mass matrix, at t
[M] : Mass matrix, at t[k] :Stiffness matrixF : Residual v
[M] [M ]U U [K] (1 )U U
ector
Ft t
),,( hvuU
Algebraic form of Sallow Water Equations
α =0 : Euler Explicit
0< α <1 : Crank–Nicolson
α=1 : Euler Implicit
![Page 18: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/18.jpg)
18
Numerical tests application of Dam Break
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
![Page 19: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/19.jpg)
19
Numerical tests application for Dam break
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Magnitude of elevation for transect 1
![Page 20: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/20.jpg)
20
Numerical tests application for Dam break
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
magnitude of elevation for transect 2
![Page 21: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/21.jpg)
21
MULTIPHYSICS-200909-11 December 2009, LILLE ,France
Any questions?
Thank youfor your attention
![Page 22: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/22.jpg)
22
M. Darbani, A. Ouahsine, P. Villon
Université de Technologie de Compiègne, laboratoire Roberval UMR-CNRS 5263, France
MULTIPHYSICS-200909-11 December 2009, LILLE ,France
Natural Elements Method for ShallowWater Equations
UTCUTC
![Page 23: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/23.jpg)
23
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication 5/5
The global matricial form reads :
)(
)1(
11
,:][
,:][
)1(][][][
n
n
nnnn
tatmatrixmassM
tatmatrixmassM
FUUKUt
MUt
M
=0 : Euler Explicite
=1 : Euler Implicite
Numerical resolution :
0( **20
dp.dfp)).(t
ρ vvvvkvvv
),,( hvuU ),,( wvuv
![Page 24: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/24.jpg)
24
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication 1/3
Shallow water equations Integration over the total water depth
0
1
1
0
0
[vH]y
[uH]xt
h
vFy
Pρy
hgfuyvv
xuu
tv
uFx
Pρx
hgfvyuv
xuu
tu
yatm
xatm
Numerical resolution
![Page 25: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/25.jpg)
25
Why meshless method ?
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Shape functions depends only on the position of nodes
In Meshless Methods the possibility of treating the problems is easier in large deformation than in the finite element method
Ability to insert or remove the nodes easily Example: Domain enrechissement
![Page 26: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/26.jpg)
26
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Shape Function of NEM (Sibsonian)
n
ixix
x
xii KK
KK
x1
,)(
)()()(
bcefAirabcdAirxi
Example
is at least C1 in every points but only
continuous at the nodal position
)(xi
![Page 27: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/27.jpg)
27
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Mathematical formulation
Cells 1st and 2d order
ijkXXdXXdXXdRxT kjim
ij ),,(),(),(:
jixxdxxdRxT jim
i ),,(),(:
The Natural Neighbors of a node are the nodes associated with neighboring Voronoi cell, or which are connected to the node by a edge of Delaunay triangle.
The first order and second order cells of the Voronoï diagram are defined mathematically by :
Natural Neighbor
![Page 28: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/28.jpg)
28
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Iterative process
ii
oldi
old vNv
3
1
)()(
t vxXDetermine vertex of triangle at t (n-1) step :
oldFinding velocity at the point with interpolation :
tv oldnew )(Find at t (n-1) step with:new
n
n
tatPtGauss
tatPtGauss
.
. 1
Adjusting the point position in the old triangle
![Page 29: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/29.jpg)
29
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
d
xN
NAdxN
ANk
ji
jik
j k
jki vv i
May be approximatted by
vanishesofoncontributithejiwhenj ,
Thus, for any arbitrary vector b we can write
dnNdN
xd
xN
N kki
ikkj k
jik b
2)b
21(
22b
0
dxN
Nk
ji onN i 0
dvxN
AN jj k
jki
![Page 30: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/30.jpg)
30
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
dxN
xNN
kkki
2i2
1i1 )v(v)v(vA
12
11 2
1,2
ik2ik NN
jiwhen
ij k
iii
k
jijj
j k
ji d
xNNd
xN
NdxN
N vvAvA kk
131
1
dNi
By taking into account of
1 2: 1
j
jj k k k
N N NPartition of unity Nx x x
![Page 31: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/31.jpg)
31
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Nodal Integration
ii
dnNdN ii
i
dnNL
N ii
appi
)(1)(
Divergence theorem :
Based on the substitution of the gradient term by an average gradient of each node in an area surrounding the representative node
Method of Stabilized Conforming Nodal Integration (SCNI)
![Page 32: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/32.jpg)
32
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
Shape Function of NEM (Sibsonian)
Example
nxi
i x xii 1x
K(x) ,K KK
3Air(cdef )(x)Air(abcd)
![Page 33: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/33.jpg)
33
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
The pressure term leads to the following integral:
dvxN
AN jj k
jki
For example: Consider the nodes 1,2,3,,..7 as natural neighbors of the node i
Pressure Term
![Page 34: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/34.jpg)
34
Solution Stability
Evolution of the particle nodes
Initial
Final
![Page 35: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/35.jpg)
35
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
221
11 )2
()2
( eeie
keeie
eki
k k
jki n
vvAn
vvAdv
xN
AN
![Page 36: M. Darbani, A. Ouahsine, P. Villon University of Technology of Compiègne](https://reader031.fdocuments.in/reader031/viewer/2022020417/56815b72550346895dc96ad5/html5/thumbnails/36.jpg)
36
Problem Shape Function SWE&Time discretizatio Nodal Integration Numerical test Appllication
The weak form of pressure term leads to the following integral:j
i k jΩj k
NN A v dΩ
x
Pressure Term
221
11 )2
()2
( eeie
keeie
eki
k k
jki n
vvAn
vvAdv
xN
AN