ERDC Talk

Outline Introduction Weighted Residuals Poisson Equation Other Examples Essential BCs Some Extensions Reference
LibMesh Experience and Usage
John W. Peterson
peterson@cfdlab.ae.utexas.edu
Univ. of Texas at Austin
January 12, 2007

1 Introduction
2 Weighted Residuals
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
1 Introduction
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
Library Description
Tumor
Angiogenesis
Compressible
RBM NS Library Structure
Basic libraries
DD are LibMesh’s
“roots”
LibMesh Application
“branches” built
Petsc off the library
Metis
Laspack “trunk”
STL
Mpich
A Generic BVP
We assume there is a ∂ΩD

Boundary Value
Problem of the form
∂u
M = F(u) ∈Ω
∂t Ω
G(u) = 0 ∈Ω
u = uD ∈ ∂ΩD
N(u) = 0 ∈ ∂ΩN ∂ΩN
u(x, 0) = u0 (x)
A Generic BVP
Associated to the
Ω
h
problem domain Ω is a
LibMesh data structure
called a Mesh
A Mesh is essentially a
collection of finite
elements
[
Ωh := Ωe
e
A Generic BVP
Associated to the
Ω
h
problem domain Ω is a
LibMesh data structure
called a Mesh
A Mesh is essentially a
collection of finite
elements
[
Ωh := Ωe
e
LibMesh provides some simple structured mesh

generation routines as well as an interface to
Triangle.
1 Introduction
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
The point of departure in any FE analysis which uses

LibMesh is the weighted residual statement
(F(u), v) = 0 ∀v ∈ V
Or, more precisely, the weighted residual statement

associated with the finite-dimensional space V h ⊂ V
(F(uh ), vh ) = 0 ∀vh ∈ V h
The point of departure in any FE analysis which uses

LibMesh is the weighted residual statement
(F(u), v) = 0 ∀v ∈ V
Or, more precisely, the weighted residual statement

associated with the finite-dimensional space V h ⊂ V
(F(uh ), vh ) = 0 ∀vh ∈ V h
Some Examples
Poisson Equation
−∆u = f ∈ Ω
Some Examples
Poisson Equation
−∆u = f ∈ Ω
Weighted Residual Statement

Z
(F(u), v) := [∇u · ∇v − fv] dx
ΩZ
+ (∇u · n) v ds
∂ΩN
Some Examples
Linear Convection-Diffusion
−k∆u + b · ∇u = f ∈ Ω
Some Examples
Linear Convection-Diffusion
−k∆u + b · ∇u = f ∈ Ω

Z
(F(u), v) := [k∇u · ∇v + (b · ∇u)v − fv] dx
Ω Z
+ k (∇u · n) v ds
∂ΩN
Some Examples
Stokes Flow
∇p − ν∆u = f
∈ Ω
∇·u = 0
Some Examples
Stokes Flow
∇p − ν∆u = f
∈ Ω
∇·u = 0

u := [u, p] , v := [v, q]
Z
(F(u), v) := [−p (∇ · v) + ν∇u :∇v − f · v
ZΩ
+ (∇ · u) q] dx + n · (ν∇u − pI) · v ds
∂ΩN
Some Examples
To obtain the approximate problem, we simply

replace u ← uh , v ← vh , and Ω ← Ωh in the weighted
residual statement.
1 Introduction
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
For simplicity we will focus on the weighted residual

statement arising from the Poisson equation, with
∂ΩN = ∅,
(F(uh ), vh ) :=
Z
h
∇u · ∇vh − fvh dx = 0 ∀vh ∈ V h

Ωh
Element Integrals
The integral over Ωh . . .
Z
h
∇u · ∇vh − fvh dx ∀vh ∈ V h

0 =
Ωh
Element Integrals
The integral over Ωh . . . is written as a sum of

integrals over the Ne finite elements:
Z
h

0 =
Ωh
Ne Z
X h

=
e=1 Ωe
Finite Element Basis Functions
An element integral will

have contributions only
from the global basis
functions corresponding
to its nodes. Ωe
We call these local basis
functions φi , 0 ≤ i ≤ Ns .
00000
11111
11111
00000
00000
11111
00000
11111
00000
11111
00000
11111
Ns
X
h

v Ωe
= ci φi
i=1
Finite Element Basis Functions
An element integral will

have contributions only
from the global basis
functions corresponding
to its nodes. Ωe
We call these local basis
functions φi , 0 ≤ i ≤ Ns .
00000
11111
11111
00000
00000
11111
00000
11111
00000
11111
00000
11111
Ns
X
h

v Ωe
= ci φi
i=1
Z Ns
X Z
h
v dx = ci φi dx
Ωe i=1 Ωe
Element Matrix and Load Vector
The element integrals . . .

Z
h
∇u · ∇vh − fvh dx

Ωe

Z
h

Ωe
are written in terms of the local “φi ” basis functions

Ns
X Z Z
uj ∇φj · ∇φi dx − f φi dx , i = 1, . . . , Ns
j=1 Ωe Ωe

Z
h

Ωe
are written in terms of the local “φi ” basis functions

Ns
X Z Z
uj ∇φj · ∇φi dx − f φi dx , i = 1, . . . , Ns
j=1 Ωe Ωe
This can be expressed naturally in matrix notation as
K e Ue − Fe
Global Linear System
The entries of the element stiffness matrix are the

integrals Z
K eij := ∇φj · ∇φi dx
Ωe

integrals Z
Ωe
While for the element right-hand side we have

Z
e
Fi := f φi dx
Ωe

integrals Z
Ωe
While for the element right-hand side we have

Z
e
Fi := f φi dx
Ωe
The element stiffness matrices and right-hand sides

can be “assembled” to obtain the global system of
equations
KU = F
Reference Element Map
The integrals are performed on a “reference” element

Ω̂e
x
Ωe
x (ξ ) ξ
Ωe

Ω̂e
x
Ωe
x (ξ ) ξ
Ωe
The Jacobian of the map x(ξ) is J.
Z Z
Fei = f φi dx = f (x(ξ))φi |J|dξ
Ωe Ω̂e

Ω̂e
x
Ωe
x (ξ ) ξ
Ωe
ˆξ
Chain rule: ∇ = J −1 ∇ξ := ∇
Z Z
K eij = ∇φj · ∇φi dx = ˆ ξ φj · ∇
∇ ˆ ξ φi |J|dξ
Ωe Ω̂e
Element Quadrature
The integrals on the “reference” element are

approximated via numerical quadrature.
Element Quadrature

The quadrature rule has Nq points “ξq ” and weights
“wq ”.
Element Quadrature

“wq ”.
Z
Fei = f φi |J|dξ
Ω̂e
Nq
X
≈ f (x(ξq ))φi (ξq )|J(ξq )|wq
q=1
Element Quadrature

“wq ”.
Z
K eij = ˆ ξ φj · ∇
∇ ˆ ξ φi |J|dξ
Ω̂e
Nq
X
≈ ˆ ξ φj (ξq ) · ∇
∇ ˆ ξ φi (ξq )|J(ξq )|wq
q=1
LibMesh Quadrature Point Data
LibMesh provides the following variables at each

quadrature point q
Code Math Description

JxW[q] |J(ξq )|wq Jacobian times weight
phi[i][q] φi (ξq ) value of ith shape fn.
ˆ ξ φi (ξq )
dphi[i][q] ∇ value of ith shape fn. gradient
xyz[q] x(ξq ) location of ξq in physical space
Matrix Assembly Loops
The LibMesh representation of the matrix and rhs

assembly is similar to the mathematical statements.
for (q=0; q<Nq; ++q)

for (i=0; i<Ns; ++i) {
Fe(i) += JxW[q]*f(xyz[q])*phi[i][q];
for (j=0; j<Ns; ++j)

Ke(i,j) += JxW[q]*(dphi[j][q]*dphi[i][q]);
}


for (i=0; i<Ns; ++i) {

}
Nq
X
Fei = f (x(ξq ))φi (ξq )|J(ξq )|wq
q=1


for (i=0; i<Ns; ++i) {

}
Nq
X
q=1


for (i=0; i<Ns; ++i) {

}
Nq
X
q=1


for (i=0; i<Ns; ++i) {

}
Nq
X
q=1


for (i=0; i<Ns; ++i) {

}
Nq
X
K eij = ˆ ξ φj (ξq ) · ∇
q=1


for (i=0; i<Ns; ++i) {

}
Nq
X
q=1


for (i=0; i<Ns; ++i) {

}
Nq
X
q=1
1 Introduction
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
Convection-Diffusion Equation
The matrix assembly routine for the linear

convection-diffusion equation,
−k∆u + b · ∇u = f

for (i=0; i<Ns; ++i) {

Ke(i,j) += JxW[q]*(k*(dphi[j][q]*dphi[i][q])
+(b*dphi[j][q])*phi[i][q]);
}

−k∆u + b · ∇u = f

for (i=0; i<Ns; ++i) {

}

−k∆u + b · ∇u = f

for (i=0; i<Ns; ++i) {

}
Stokes Flow
For multi-variable systems like Stokes flow,
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
The element stiffness matrix concept can extended to
include sub-matrices
 e
Ku1 u1 Kue1 u2 Kue1 p
 e   e 
Uu1 Fu1
 Kue u Kue u Kue p   Uue  −  Fue 
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe
We have an array of submatrices: Ke[ ][ ]

Stokes Flow
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
 e
 e   e 
Uu1 Fu1
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe
We have an array of submatrices: Ke[0][0]

Stokes Flow
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
 e
 e   e 
Uu1 Fu1
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe

Stokes Flow
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
 e
 e   e 
Uu1 Fu1
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe

Stokes Flow
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
 e
 e   e 
Uu1 Fu1
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe
And an array of right-hand sides: Fe[].

Stokes Flow
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
 e
 e   e 
Uu1 Fu1
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe
And an array of right-hand sides: Fe[0].

Stokes Flow
∇p − ν∆u = f
∈ Ω ⊂ R2
∇·u = 0
 e
 e   e 
Uu1 Fu1
2 1 2 2 2 2 2
e
Kpu 1
Kpue
2
Kppe
Upe Fpe
And an array of right-hand sides: Fe[1].

Stokes Flow
The matrix assembly can proceed in essentially the

same way.
For the momentum equations:

for (d=0; d<2; ++d)
for (i=0; i<Ns; ++i) {
Fe[d](i) += JxW[q]*f(xyz[q],d)*phi[i][q];

Ke[d][d](i,j) +=
JxW[q]*nu*(dphi[j][q]*dphi[i][q]);
}
1 Introduction
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
Essential Boundary Data
Dirichlet boundary conditions can be enforced after

the global stiffness matrix K has been assembled
This usually involves
1 placing a “1” on the main diagonal of the global
stiffness matrix
2 zeroing out the row entries
3 placing the Dirichlet value in the rhs vector
4 subtracting off the column entries from the rhs

stiffness matrix

stiffness matrix

stiffness matrix

stiffness matrix
       
k11 k12 k13 . f1 1 0 0 0 g1
k21 k22 k23 . f2 
   0 k22 k23 . , f2 − k21 g1 
 
 , →
k31 k32 k33 . f3  0 k32 k33 .  f3 − k31 g1 
. . . . . 0 . . . .
Cons of this approach :

Works for an interpolary finite element basis but not in
general.
May be inefficient to change individual entries once
the global matrix is assembled.
Need to enforce boundary conditions for a generic
finite element basis at the element stiffness matrix
level.

general.
level.

general.
level.

general.
level.
Penalty Formulation
One solution is the “penalty” boundary formulation

A term is added to the standard weighted residual
statement
Z
1
(F(u), v) + (u − uD )v dx = 0 ∀v ∈ V
∂ΩD
| {z }
penalty term
Here 1 is chosen so that, in floating point

arithmetic, 1 + 1 = 1 .
This weakly enforces u = uD on the Dirichlet
boundary, and works for general finite element bases.
Penalty Formulation

statement
Z
1
(F(u), v) + (u − uD )v dx = 0 ∀v ∈ V
∂ΩD
| {z }
penalty term

Penalty Formulation

statement
Z
1
(F(u), v) + (u − uD )v dx = 0 ∀v ∈ V
∂ΩD
| {z }
penalty term

Penalty Formulation

statement
Z
1
(F(u), v) + (u − uD )v dx = 0 ∀v ∈ V
∂ΩD
| {z }
penalty term

Penalty Formulation
LibMesh provides:
A quadrature rule with Nqf points and JxW f[]
A finite element coincident with the boundary face
that has shape function values phi f[][]
for (qf=0; qf<Nqf; ++qf) {

for (i=0; i<Nf; ++i) {
Fe(i) += JxW_f[qf]*
penalty*uD(xyz[q])*phi_f[i][qf];
for (j=0; j<Nf; ++j)

Ke(i,j) += JxW_f[qf]*
penalty*phi_f[j][qf]*phi_f[i][qf];
}
}
Penalty Formulation
LibMesh provides:

for (i=0; i<Nf; ++i) {
Fe(i) += JxW_f[qf]*

}
}
Penalty Formulation
LibMesh provides:

for (i=0; i<Nf; ++i) {
Fe(i) += JxW_f[qf]*

}
}
1 Introduction
3 Poisson Equation
4 Other Examples
5 Essential BCs
6 Some Extensions
Time-Dependent Problems
For linear problems, we have already seen how the

weighted residual statement leads directly to a
sparse linear system of equations
KU = F
Time-Dependent Problems
For time-dependent problems,

∂u
= F(u)
∂t
we also need a way to advance the solution in time,
e.g. a θ-method
n+1
u − un h

= F(uθ ), vh ∀vh ∈ V h

,v
∆t
uθ := θun+1 + (1 − θ)un
Leads to KU = F at each timestep.

Nonlinear Problems
For nonlinear problems, typically a sequence of linear

problems must be solved, e.g. for Newton’s method
(F 0 (uk )δuk+1 , v) = −(F(uk ), v)
where F 0 (uk ) is the linearized (Jacobian) operator

associated with the PDE.
Must solve KU = F (Inexact Newton method) at each
iteration step.
Non-Trivial Applications
Common to each is the need to solve a linear system

(or systems) of equations KU = F.
LibMesh provides several of the tools necessary to
construct these systems, but it is not specifically
written to solve any one problem.
We now briefly show a few of the non-trivial
applications which have been built on top of the
library.
Natural Convection
Tetrahedral mesh of “pipe” geometry. Stream ribbons

colored by temperature.
Surface-Tension-Driven Flow
Adaptive grid solution shown with temperature

contours and velocity vectors.
Double-Diffusive Convection
Solute contours: a plume of warm, low-salinity fluid is

convected upward through a porous medium.
Tumor Angiogenesis
The tumor secretes a chemical which stimulates

blood vessel formation.
Natural Convection
Tetrahedral mesh of “pipe” geometry. Stream ribbons

colored by temperature.
Surface-Tension-Driven Flow
Adaptive grid solution shown with temperature

contours and velocity vectors.
Double-Diffusive Convection
Solute contours: a plume of warm, low-salinity fluid is

convected upward through a porous medium.
Tumor Angiogenesis
The tumor secretes a chemical which stimulates

blood vessel formation.
B. Kirk, J. Peterson, R. Stogner and G. Carey,

“libMesh: a C++ library for parallel adaptive mesh
refinement/coarsening simulations”, Engineering with
Computers, vol. 22, no. 3–4, p. 237–254, 2006.

ERDC Talk

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

ERDC Talk

Enviado por

Direitos autorais:

Formatos disponíveis

Outline Introduction Weighted Residuals Poisson Equation Other Examples Essential BCs Some Extensions Reference

LibMesh Experience and Usage

Univ. of Texas at Austin

January 12, 2007

We assume there is a ∂ΩD

LibMesh provides some simple structured mesh

The point of departure in any FE analysis which uses

Or, more precisely, the weighted residual statement

The point of departure in any FE analysis which uses

Or, more precisely, the weighted residual statement

Weighted Residual Statement

Weighted Residual Statement

Weighted Residual Statement

To obtain the approximate problem, we simply

Weighted Residual Statement

For simplicity we will focus on the weighted residual

The integral over Ωh . . .

The integral over Ωh . . . is written as a sum of

Finite Element Basis Functions

An element integral will

Finite Element Basis Functions

An element integral will

Element Matrix and Load Vector

The element integrals . . .

Element Matrix and Load Vector

The element integrals . . .

are written in terms of the local “φi ” basis functions

Element Matrix and Load Vector

The element integrals . . .

are written in terms of the local “φi ” basis functions

This can be expressed naturally in matrix notation as

Global Linear System

The entries of the element stiffness matrix are the

Global Linear System

The entries of the element stiffness matrix are the

While for the element right-hand side we have

Global Linear System

The entries of the element stiffness matrix are the

While for the element right-hand side we have

The element stiffness matrices and right-hand sides

Reference Element Map

The integrals are performed on a “reference” element

Reference Element Map

The integrals are performed on a “reference” element

Reference Element Map

The integrals are performed on a “reference” element

The integrals on the “reference” element are

The integrals on the “reference” element are

The integrals on the “reference” element are

The integrals on the “reference” element are

LibMesh Quadrature Point Data

LibMesh provides the following variables at each

Code Math Description

Matrix Assembly Loops

The LibMesh representation of the matrix and rhs

for (q=0; q<Nq; ++q)

for (j=0; j<Ns; ++j)

Matrix Assembly Loops

The LibMesh representation of the matrix and rhs

for (q=0; q<Nq; ++q)

for (j=0; j<Ns; ++j)

Matrix Assembly Loops

The LibMesh representation of the matrix and rhs

Here 1 is chosen so that, in floating point

Here 1 is chosen so that, in floating point