midterm review: mimetic isogemetric fem - tu...
TRANSCRIPT
2
Objective:
Combine ideas from isogeometric analysis and
mimetic methods to develop a structure-preserving
discretization for the Euler equations for
incompressible fluids.
4
Project outline (cntβd)
β’ Phase I questions:
β How can we use IGA to solve PDEβs?
β What structures are facilitated in elliptic PDEβs?
β How can we preserve these structures?
β Can we construct a MIMIGA method to discretize an
elliptic PDE problem?
5
This presentation - literature review
β’ Introduction
β Isogeometric Analysis & Mimetic Methods
β’ Approach for elliptic PDEβs
β Exterior calculus
β DeRham complex
β Application: Scalar Poisson equation in 2D
β’ Conclusion
β’ Future work
6
Introduction β Isogeometric Analysis
β’ Introduced by the Hughes group in 2005 to
bridge the gap between CAD and FEM
β’ Isogeometric paradigm
β’ B-splines make an excellent basis for FEM
7
Introduction β Mimetic Methods
β’ PDEβs facilitate physical structures and
symmetries.
β’ Tools from exterior calculus and algebraic
topology are used to capture these structures.
β’ Growing awareness: Disrete exterior calculus,
discrete hodge theory, exterior finite element
method, compatible methods, mimetic finite
diference, etc
8
Why exterior calculus?
β’ Structures become apparent.
β’ Distinction between topological and metric
dependencies.
β’ Generalized for π dimensions.
9
Differential Forms; πΌ(π)
β’ Differential forms are elements from the dual
vector space,
β’ Associated with geometric structure,
β 0-form: π(0) = π π₯, π¦
β 1-form: πΌ(1) = πΌ1 π₯, π¦ ππ₯ + πΌ2 π₯, π¦ ππ¦
β’ βMeasurement of physical variables,β
β π = π(2) = π π₯, π¦ ππ₯Λππ¦
β’ Space of k-forms: Ξ(π)
10
Exterior derivative; d
β’ Exterior derivative d generalizes π»π, π» Γ π, π» β π£
ππΌ(1) =ππΌ2ππ₯
βππΌ1ππ¦
ππ₯Λππ¦
β’ π: Ξ(π) β Ξ(π+1)
β’ Exact sequence, the DeRham complex
β’ Nilpotent, πππΌ(π) = 0
β’ Independent of metric
11
Hodge-β operator;
β’ Maps forms to dual geometry,
β’ Metric dependent,
β’ Double DeRham complex,
12
Codifferential; πβ
β’ πβ ββ π β
β’ Adjoint of π: β, πβ β = π β,β - β« bcβs
β’ Laplace operator: β= ππβ + πβπ
πβπβ
13
Scalar Poisson equation
β’ E.g. Potential flow, electrostatics,
β’ Given π π₯, π¦ = 2π2 sin ππ₯ sin ππ¦find π π₯, π¦ such that βπ = π on Ξ© = [0,1]2
with π = 0 on πΞ©
0-form,
Find π(0) s.t. πβππ(0) = π(0)
2-form,
Find π(2) s.t. ππβπ(2) = π(2)
Same solution, different discretization
14
0-form Poisson; πβππ(0) = π(0)
β’ Weak formulation,π€(0), πβππ(0)
Ξ©= π€(0), π(0)
Ξ©
ππ€(0), ππ(0)Ξ©= π€(0), π(0)
Ξ©β
πΞ©
π€ 0 β§βππ 0
β’ Well-posedness through Lax-Milgram,
15
0-form Poisson; FEM
β’ Conforming FEM, take Ξ βπβ Ξ(π)
β’ Use B-spline spans Ξ β0= ππ,π
n=10, p =3
16
0-form Poisson; edge functions
β’ Applying the exterior derivative (1D-example)
β Nodal basis: πβ(0)
= π=0π ππβπ
π(π₯) = π
ππ 0
β Then, ππβ(0)
= π=1π ππ β ππβ1 ππ
πβ1(π₯) = πΌ(10)π
ππ 1
Differences of coefficientsare captured in matrix using -1,0,1
New edge type basis functionemerges with a polynomialdegree less
17
0-form Poisson; edge functions (cntβd)
n-1=9, p-1 =2
β’ Extension to 2D using tensor products of nodal and edge
type basis
β’ Nodal/edge
β 0-form
β 1-form
β 2-form
18
0-form Poisson, Matrices
β’ ππ€β(0), ππβ
(0)
Ξ©= π€π πΌ10 π Ξ© π (1)
ππ (1) πΌ10 π
β’ Result: πΌ(10)ππ(11)πΌ(10)π = π
Mass matrix π(11)
19
0-form Poisson, Matrices (cntβd)
β’ Exact discretization of π£(1) = ππ(0) through
incidence matrices, π£ = πΌ(10)π
β’ Incidence matrices are nilpotent πΌ(21)πΌ(10) = β ,
and satisfy the DeRham sequence
β’ Hodge-β operator (metric) is discretized through
mass matrix π(11)
22
2-form Poisson; ππβπ(2) = π(2)
β’ Weak formulation;
π€(2), ππβπ(2)Ξ©= π€(2), π(2)
Ξ©
β’ Integration by parts? No, take mixed formulation:
πβπ(2) = Ο(1)
πΟ(1) = π(2)
β’ Weak form:
ππ(1), π(2)Ξ©= π(1), Ο(1)
Ξ©β
πΞ©
π 1 β§βπ 2
π€(2), πΟ(1)Ξ©= π€(2), π(2)
Ξ©
β’ Well posedness through Inf-Sup conditions
23
2-form Poisson; FEM
β’ Can we take,
β Ξ β1= ππ,π?
β Ξ β2= ππ,π?
β’ No, well-posedness depends on the DeRhamsequence. We take
β Ξ β1= ππβ1,π Γ ππ,πβ1
β Ξ β2= ππβ1,πβ1
β’ Which satisfy exact sequence
d
24
2-form Poisson; Matrices
β π 1 , Ο 1
Ξ©+ ππ(1), π(2)
Ξ©= 0
π€(2), πΟ(1)Ξ©= π€(2), π(2)
Ξ©
βπ(11) π(22)πΌ(21)π
π(22)πΌ(21) β
Ο
π=
0
π
β’ Or π(22)πΌ(21)ππ(11) β1
π(22)πΌ(21) Ο = π
26
Conclusion
β’ Elliptic problems can be discretized using mass
matrices and incidence matrices.
β’ Solution spaces are chosen such that they satisfy
the DeRham complex.
27
Conclusion (cntβd)
β’ Comparison 0-form & 2-form Poisson:
0-form 2-form
πΌ(10)ππ(11)πΌ(10) βπ(11) π(22)πΌ(21)
π
π(22)πΌ(21) β
Obtain solution π(0) Obtain solutions π(2), Ο(1)
- Dirichlet is essential- Neumann is natural
- Dirichlet is natural- Neumann is essential
Gradient exact πΌ(10)π = π£ Divergence exact πΌ(21)Ο = 0
i.e. π» β π£ = 0
28
Future Work
β’ Towards the incompressible Euler equations:
β Extend to hyperbolic problems,
β Linear advection equation.
β Construction of periodic domain.
β Staggering velocity and vorticity in time?