An Introduction To Isogeometric Analysis: A. Buffa, G. Sangalli, R. V Azquez

An introduction to
Isogeometric Analysis
A. Buffa1, G. Sangalli1,2, R. Vázquez1
1 Istituto di Matematica Applicata e Tecnologie Informatiche -

Pavia
Consiglio Nazionale delle Ricerche
2 Dipartimento di Matematica, Università di Pavia
Supported by the ERC Starting Grant:

GeoPDEs n. 205004
R. Vázquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 1 / 33

1 Anoverview on IsoGeometric analysis (IGA)
B-splines and NURBS
Geometry description
Discretization in IGA
Local refinement: T-splines
2 Approximation of vector fields and differential forms

Construction of the discrete spaces
The commuting De Rham diagram
Maxwell eigenproblem: B-splines discretization
Maxwell eigenproblem: NURBS discretization

Part I
An overview on Isogeometric
Analysis

IsoGeometric Analysis (IGA): an overview
Geometry is defined by Computer Aided Design (CAD) software.

CAD is based on Non Uniform Rational B-Splines (NURBS).


CAD and FEM usedifferent descriptionsfor the geometry.

) Iso-parametric description of the geometry.
) Updating the geometry requires interface with CAD and remeshing.


CAD and FEM usedifferent descriptionsfor the geometry.

CAD and IGA use thesame geometry description.
) Maintain the geometric description given by CAD
(NURBS).
) Iso-parametric approach: PDEs are numerically solved with
NURBS.
IsoGeometric Analysis: definition of B-splines
Let {ξ1, . . . , ξn+p+1} be a non-uniform knot vector in the interval [0,1].

B-spline basis functions are defined recursively as
.
1 if ξ i ≤ ξ < ξi +1
B i(ξ)
, = ,0 otherwise.
0
x − ξi ξi +p+1 − x
B i , (x ) = Bi ,p−1(x ) + Bi +1,p−1 (x )
p ξi +p − ξi ξi +p+1 − ξi +1 .


The main properties of B-splines basis functions are
Piecewise polynomials of degree p, and regularity at most p − 1. The
regularity can be controlled by changing the knots multiplicity. The
function Bi,p is supported in the interval [ξi , ξi +p+1].
They are non-negative and form a partition of unity.


The main properties of B-splines basis functions are
Piecewise polynomials of degree p, and regularity at most p − 1. The
regularity can be controlled by changing the knots multiplicity. The
function Bi,p is supported in the interval [ξi , ξi +p+1].
They are non-negative and form a partition of unity.
S pα : space of B-splines of degree p and regularity α at the knots.
. Σ
d
Their derivatives satisfy v : v ∈ S αp ≡ Sαp−− 11
dx
.
B-spline curves in Rd
Σ
F(x ) = im=1 Bi,p (x )Ci , Ci ∈ Rd are thecontrol points.

B-spline curves in Rd
Σ
F(x ) = im=1 Bi,p (x )Ci , Ci ∈ Rd are thecontrol points.
Multivariate B-splines
The definition is generalized by tensor products:
p p p
S pα11,p,α2 ,p2 ,α
3
3
:= S α11 ⊗ Sα22 ⊗ Sα33 , Bijk (x) := Bi,p1 (x )Bj,p2 (y )Bk,p3 (z ).
B-spline volumes in R3
m1,m
Σ2,m3
F(x) = Bijk (x)Cijk , Cijk ∈ R3 are thecontrol points.
i ,j ,k=1
IsoGeometric Analysis: definition of NURBS
NURBS in Rd are conic projections of B-splines in Rd +1
Weights, control points and basis functions:

(Cw )j Bi ,p (ξ)wi .
wi = (Ci wi)d +1, (Ci )j = , Ri , = Σ m
wi p A=1 B A,p (ξ)w
A

wi = (Cw )d +1, (Ci )j = (Cwi )j Bi ,p (ξ)wi .

i
wi , Ri , = Σ
p
m
A=1 B A,p (ξ)w
A
Represent exactly a wide class of curves as, e.g., conic sections.

They have most of the good properties of B-splines.

wi = (Cw )d +1, (Ci )j = (Cwi )j Bi ,p (ξ)wi .

i
wi , Ri , = Σ
p
m
A=1 B A,p (ξ)w
A
Represent exactly a wide class of curves as, e.g., conic sections.

They have most of the good properties of B-splines.
N pα : space of NURBS of degree p and regularity α at the knots.
IGA: geometry description and mesh refinement
Coarsest mesh: geometry description
Physical domain Ω
“Patch”, Ω^ = (0, 1)2
, , ,. ,
w iBi w i Bi
Ri = w Σ w
◦ F− 1
i =1,...,N0 i =1,...,N0
The approximation space Vh on Ω is obtained by push-forward:

Vh = span{Ri ◦ F−1 , i = 1, . . . , N0}
First refinement
“Patch”, Ω = (0, 1)2 Physical domain Ω

^
F
, , ,. ,
w iBi w i Bi
Ri = w Σ w
◦ F− 1
i =1,...,N1 i =1,...,N1
The geometrical map F and the weight w arefixedat the coarsest level of
discretization!

Second refinement... and so on
Physical domain Ω
“Patch”, Ω^ = (0, 1)2
, , ,. ,
w iBi w i Bi
Ri = w Σ w
◦ F− 1
i =1,...,N2 i =1,...,N2
The main drawback is thetensor product structure.

Remark on refinement
There are three possibilities for refinement:

h-refinement (by multiple knot insertion)


p-refinement (by degree elevation)


p-refinement (by degree elevation)
k-refinement (by degree and continuity elevation)

Multipatch domains
CAD : Geometries are described by mappings of several patches.
F1
1
1
F2 2
 3
0 1
F3
Patch interfaces are normally treated just imposing C 0 regularity

⇓
domain decomposition type structure
Breaking the tensor product structure: T-splines
Sederberg et al. 2004-, Buffa, Cho, Sangalli et al. 2010
CAD community (2004-): Definition ofT-splines.

Based on PB splines, are associated with a T-mesh.
They are not based on tensor product structure.






Courtesy of Sederberg et al 2004


Refinement algorithm ensures S h⊆ S ref (Seberberg et al )

h
) Possible severe fill-in of the T-mesh
) Expensive (cycle on many elements) and not local
) There is no well defined de-refinement strategy!


Refinement algorithm ensures S h⊆ S ref (Seberberg et al )

h
) Possible severe fill-in of the T-mesh
) Expensive (cycle on many elements) and not local
) There is no well defined de-refinement strategy!
Our contributions:
) Locality analysis: C 2 versus C 1 cubic splines
) Linear independence for fairly general meshes

Severe fill-in: the worst case scenario
Want to refine the gray quads by split them in 4

Severe fill-in: the worst case scenario
Want to refine the gray quads C 2 basis functions allover

Severe fill-in can be avoided

Severe fill-in: second refinement step

Severe fill-in: second refinement step

Continuity in the use of T-splines
Buffa, Kumar, Sangalli. In preparation
The same behavior is observed in numerical simulations.
C 2 continuity causes a
fill-in of the mesh.

Continuity in the use of T-splines
Buffa, Kumar, Sangalli. In preparation
The same behavior is observed in numerical simulations.
C 2 continuity causes a
fill-in of the mesh.
The refinement
with
C 1 continuity is local.

Linearly Independent T-splines
Buffa, Cho, Sangalli, 2010
There are examples of linearly dependent T-splines.

In cases of interest, we have a result. Indeed,linear independence is
guaranteed for all refinements obtained by the local refinement
algorithm and that can be decomposed on successive insertion of new
lines.


lines.
E.g.,


lines.
E.g.,


lines.
E.g.,

Open issues
Local or quasi local refinement algorithm, allowing for regular

functions.
The definition of “aligned” T-splines may help.
Hughes, Scott et al. In preparation.

Open issues

functions.
Linear independence on general T-meshes.

Open issues

functions.
Control on support intersection.

Open issues

functions.
Local approximation properties.
Beiraõ da Veiga, Buffa, Cho, Sangalli. In preparation.

Open issues

functions.
Suitable derefinement strategy.

Open issues

functions.
Suitable derefinement strategy.
Some of these problems may be
solved with LR-splines (locally
refined).
Dokken, Lyche et al. In preparation

An Introduction To Isogeometric Analysis: A. Buffa, G. Sangalli, R. V Azquez

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

An Introduction To Isogeometric Analysis: A. Buffa, G. Sangalli, R. V Azquez

Uploaded by

Copyright:

Available Formats

An introduction to

A. Buffa1, G. Sangalli1,2, R. V´azquez1

1 Istituto di Matematica Applicata e Tecnologie Informatiche -

Supported by the ERC Starting Grant:

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 1 / 33

2 Approximation of vector fields and differential forms

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 2 / 33

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 3 / 33

Geometry is defined by Computer Aided Design (CAD) software.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 4 / 33

Geometry is defined by Computer Aided Design (CAD) software.

CAD and FEM usedifferent descriptionsfor the geometry.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 4 / 33

Geometry is defined by Computer Aided Design (CAD) software.

CAD and FEM usedifferent descriptionsfor the geometry.

Let {ξ1, . . . , ξn+p+1} be a non-uniform knot vector in the interval [0,1].

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 5 / 33

Let {ξ1, . . . , ξn+p+1} be a non-uniform knot vector in the interval [0,1].

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 5 / 33

Let {ξ1, . . . , ξn+p+1} be a non-uniform knot vector in the interval [0,1].

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 5 / 33

Weights, control points and basis functions:

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 6 / 33

Weights, control points and basis functions:

wi = (Cw )d +1, (Ci )j = (Cwi )j Bi ,p (ξ)wi .

Represent exactly a wide class of curves as, e.g., conic sections.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 6 / 33

Weights, control points and basis functions:

wi = (Cw )d +1, (Ci )j = (Cwi )j Bi ,p (ξ)wi .

Represent exactly a wide class of curves as, e.g., conic sections.

Coarsest mesh: geometry description

The approximation space Vh on Ω is obtained by push-forward:

“Patch”, Ω = (0, 1)2 Physical domain Ω

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 7 / 33

Second refinement... and so on

The main drawback is thetensor product structure.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 7 / 33

There are three possibilities for refinement:

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 8 / 33

There are three possibilities for refinement:

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 8 / 33

There are three possibilities for refinement:

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 8 / 33

Patch interfaces are normally treated just imposing C 0 regularity

CAD community (2004-): Definition ofT-splines.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 10 / 33

CAD community (2004-): Definition ofT-splines.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 10 / 33

CAD community (2004-): Definition ofT-splines.

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 10 / 33

CAD community (2004-): Definition ofT-splines.

Courtesy of Sederberg et al 2004

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 10 / 33

CAD community (2004-): Definition ofT-splines.

Refinement algorithm ensures S h⊆ S ref (Seberberg et al )

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 10 / 33

CAD community (2004-): Definition ofT-splines.

Refinement algorithm ensures S h⊆ S ref (Seberberg et al )

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 10 / 33

Want to refine the gray quads by split them in 4

R. V´azquez (IMATI-CNR Italy) Introduction to Isogeometric Analysis Santiago de Compostela, 2010 11 / 33

Want to refine the gray quads C 2 basis functions allover