BAIL 2006 Book of Abstracts - Institut für Numerische und ...
BAIL 2006 Book of Abstracts - Institut für Numerische und ... BAIL 2006 Book of Abstracts - Institut für Numerische und ...
R. SCHNEIDER, P. JIMACK: Anisotropic mesh adaption based on a posteriori estimates and optimisation of node positions ✬ ✫ Anisotropic mesh adaption based on a posteriori estimates and optimisation of node positions Abstract Introduction René Schneider ∗ & Peter Jimack † Efficient numerical approximation of solution features like boundary or interior layers by means of the finite element method requires the use of layer adapted meshes. Anisotropic meshes, like for example Shishkin meshes, allow the most efficient approximation of these highly anisotropic solution features. However, application of this approach relies on a priori analysis on the thickness, position and stretching direction of the layers. If it is impossible to obtain this information a priori, as it is often the case for problems with interior layers of unknown position for example, automatic mesh adaption based on a posteriori error estimates or error indicators is essential in order to obtain efficient numerical approximations. Historically the majority of work on automatic mesh adaption used locally uniform refinement, splitting each element into smaller elements of similar shape. This procedure is clearly not suitable to produce anisotropically refined meshes. The resulting meshes are over-refined in at least one spatial direction, rendering the approach far less efficient than that of the anisotropic meshes based on a priori analysis. Automatic anisotropic mesh adaption is an area of active research, e.g. [2, 1]. Here we present a new approach to this problem, based upon using not only an a posteriori error estimate to guide the mesh refinement, but its sensitivities with respect to the positions of the nodes in the mesh as well. Outline of the approach The underlying idea is to use techniques from mathematical optimisation to minimise the estimated error by moving the positions of the nodes in the mesh appropriately. This basic idea is of course not new, but the approach taken to realise it is. A key ingredient is the utilisation of the discrete adjoint technique to evaluate the sensitivities of an error estimate J = J(uh(s),s) with respect to the node positions s, where uh = uh(s) denotes the solution of the discretised PDE, R(uh,s) = 0, which depends upon the node positions s. The sensitivities � ∂R ∂uh DJ ∂J ∂uh ∂J = + Ds ∂uh ∂s ∂s � T ∗ Chemnitz University of Technology, Germany † School of Computing, University of Leeds, UK = ∂J ∂s ∂R − ΨT , (1) ∂s Ψ = ∂J , (2) ∂uh Speaker: SCHNEIDER, R. 28 BAIL 2006 ✩ ✪
R. SCHNEIDER, P. JIMACK: Anisotropic mesh adaption based on a posteriori estimates and optimisation of node positions ✬ ✫ y 1 0.8 0.6 0.4 0.2 0 −0.2 −0.4 −0.6 −0.8 −1 initial mesh −1 −0.5 0 x 0.5 1 y 1 0.8 0.6 0.4 0.2 0 −0.2 −0.4 −0.6 −0.8 −1 final mesh for eps=1.0e−03 −1 −0.5 0 x 0.5 1 (a) (b) Figure 1: Initial (a) and adapted (b) meshes for model problem (3) with a boundary layer around the square hole are thereby evaluated according to (1), utilising the adjoint solution Ψ which is defined by (2). This way, DJ/Ds can be evaluated without computing ∂uh/∂s first, reducing the number of equation systems to be solved from O(dim(s)) to just two, independent of dim(s). As the number of nodes can easily be larger than one hundred even in extremely coarse meshes (if the domain geometry is complicated), this approach is of fundamental importance to make gradient based optimisation methods feasible for this type of problem. Fast optimisation algorithms like BFGS-type methods can be applied to obtain significant reductions in the error estimate J after a few optimisation steps. The aim of this procedure is to provide a mesh with problem adapted anisotropic elements, which may then be used as a good basis for further locally uniform adaptive mesh refinement. Numerical results To demonstrate feasibility of the approach it is applied to a number of model problems. For the purpose of this abstract we consider a reaction diffusion equation, −∆u + 1 ε2u = 1 ε2 in Ω := (−1,1) 2 \ (−1 1 5 , 5 )2 subject to u = 0 on ΓD := � −1 ⎫ ⎪⎬ � 1 2 � � 1 1 2 5 , 5 \ −5 , (3) 5 ⎪⎭ ∂u ∂n = 0 on ΓN := [−1,1] 2 \ (−1,1) 2 . Figure 1 shows an initial mesh for this problem and an adapted one for ε = 10 −3 . Concentration of the elements in the boundary layer which forms around the hole at the centre of the domain is clearly visible, and significantly increased aspect ratios in the boundary layer may be observed. For more detail on the approach and the example we refer to [3]. References [1] L. Formaggia, S. Micheletti, and S. Perotto. Anisotropic mesh adaptation in computational fluid dynamics: Application to the advection-diffusion-reaction and the Stokes problems. Applied Numerical Mathematics, 51(4):511–533, December 2004. [2] G. Kunert. Toward anisotropic mesh construction and error estimation in the finite element method. Numerical Methods for Partial Differential Equations, 18(5):625–648, 2002. [3] R. Schneider and P.K. Jimack. Toward anisotropic mesh adaption based upon sensitivity of a posteriori estimates. School of Computing Research Report Series 2005.03, University of Leeds, http://www. comp.leeds.ac.uk/research/pubs/reports/2005/2005_03.pdf, 2005. Speaker: SCHNEIDER, R. 29 BAIL 2006 ✩ ✪
- Page 1: BAIL 2006 International Conference
- Page 5 and 6: Plenary Session 1 Session 2 Session
- Page 7 and 8: Room School-Lab (SL) Room MPI Sessi
- Page 9 and 10: Room School-Lab (SL) Room MPI Tuesd
- Page 11 and 12: Room School-Lab (SL) Room MPI Sessi
- Page 13 and 14: Room School-Lab (SL) Room MPI Sessi
- Page 15 and 16: Room School-Lab (SL) Room MPI Minis
- Page 17 and 18: Contents Greetings . . . . . . . .
- Page 19 and 20: CONTENTS S. HEMAVATHI, S. VALARMATH
- Page 21 and 22: CONTENTS G. LUBE: A stabilized fini
- Page 23: Plenary Presentations
- Page 26 and 27: P. HOUSTON: Discontinuous Galerkin
- Page 28 and 29: M. STYNES: Convection-diffusion pro
- Page 30 and 31: V. GRAVEMEIER, S. LENZ, W.A. WALL:
- Page 33: Minisymposia
- Page 36 and 37: R.K. DUNNE, E. O’RIORDAN, M.M. TU
- Page 38 and 39: G.I. SHISHKIN: A posteriori adapted
- Page 40 and 41: W. LAYTON, I. STANCULESCU: Numerica
- Page 42 and 43: H. WANG: A Component-Based Eulerian
- Page 44 and 45: R. HARTMANN: Discontinuous Galerkin
- Page 46 and 47: V. HEUVELINE: On a new refinement s
- Page 48 and 49: J.A. MACKENZIE, A. NICOLA: A Discon
- Page 53 and 54: Speaker: Debopam Das, Tapan Sengupt
- Page 55 and 56: M.H. BUSCHMANN, M. GAD-EL-HAK: Turb
- Page 57 and 58: A. NAYAK, D. DAS: Three-dimesnional
- Page 59 and 60: J. HUSSONG, N. BLEIER, V.V. RAM: Th
- Page 61 and 62: T.K. SENGUPTA, A. KAMESWARA RAO: Sp
- Page 63 and 64: L. SAVIĆ, H. STEINRÜCK: Asymptoti
- Page 65 and 66: G.I. Shiskin, P. Hemker Robust Meth
- Page 67 and 68: D. BRANLEY, A.F. HEGARTY, H. PURTIL
- Page 69 and 70: T. LINSS, N. MADDEN: Layer-adapted
- Page 71 and 72: L.P. SHISHKINA, G.I. SHISHKIN: A Di
- Page 73 and 74: I.V. TSELISHCHEVA, G.I. SHISHKIN: D
- Page 75 and 76: S. HEMAVATHI, S. VALARMATHI: A para
- Page 77 and 78: J. Maubach, I, Tselishcheva Robust
- Page 79 and 80: M. ANTHONISSEN, I. SEDYKH, J. MAUBA
- Page 81 and 82: S. LI, L.P. SHISHKINA, G.I. SHISHKI
- Page 83 and 84: A.I. ZADORIN: Numerical Method for
- Page 85 and 86: P. ZEGELING: An Adaptive Grid Metho
- Page 87: Contributed Presentations
- Page 90 and 91: F. ALIZARD, J.-CH. ROBINET: Two-dim
- Page 92 and 93: TH. ALRUTZ, T. KNOPP: Near-wall gri
- Page 94 and 95: M. BAUSE: Aspects of SUPG/PSPG and
- Page 96 and 97: L. BOGUSLAWSKI: Sheare Stress Distr
- Page 98 and 99: A.CANGIANI, E.H.GEORGOULIS, M. JENS
R. SCHNEIDER, P. JIMACK: Anisotropic mesh adaption based on a posteriori estimates<br />
and optimisation <strong>of</strong> node positions<br />
✬<br />
✫<br />
Anisotropic mesh adaption based on a posteriori estimates and<br />
optimisation <strong>of</strong> node positions<br />
Abstract<br />
Introduction<br />
René Schneider ∗ & Peter Jimack †<br />
Efficient numerical approximation <strong>of</strong> solution features like bo<strong>und</strong>ary or interior layers by means<br />
<strong>of</strong> the finite element method requires the use <strong>of</strong> layer adapted meshes. Anisotropic meshes, like<br />
for example Shishkin meshes, allow the most efficient approximation <strong>of</strong> these highly anisotropic<br />
solution features. However, application <strong>of</strong> this approach relies on a priori analysis on the thickness,<br />
position and stretching direction <strong>of</strong> the layers. If it is impossible to obtain this information<br />
a priori, as it is <strong>of</strong>ten the case for problems with interior layers <strong>of</strong> unknown position for example,<br />
automatic mesh adaption based on a posteriori error estimates or error indicators is essential in<br />
order to obtain efficient numerical approximations.<br />
Historically the majority <strong>of</strong> work on automatic mesh adaption used locally uniform refinement,<br />
splitting each element into smaller elements <strong>of</strong> similar shape. This procedure is clearly not<br />
suitable to produce anisotropically refined meshes. The resulting meshes are over-refined in at<br />
least one spatial direction, rendering the approach far less efficient than that <strong>of</strong> the anisotropic<br />
meshes based on a priori analysis.<br />
Automatic anisotropic mesh adaption is an area <strong>of</strong> active research, e.g. [2, 1]. Here we present<br />
a new approach to this problem, based upon using not only an a posteriori error estimate to<br />
guide the mesh refinement, but its sensitivities with respect to the positions <strong>of</strong> the nodes in the<br />
mesh as well.<br />
Outline <strong>of</strong> the approach<br />
The <strong>und</strong>erlying idea is to use techniques from mathematical optimisation to minimise the estimated<br />
error by moving the positions <strong>of</strong> the nodes in the mesh appropriately. This basic idea is<br />
<strong>of</strong> course not new, but the approach taken to realise it is.<br />
A key ingredient is the utilisation <strong>of</strong> the discrete adjoint technique to evaluate the sensitivities<br />
<strong>of</strong> an error estimate J = J(uh(s),s) with respect to the node positions s, where uh = uh(s)<br />
denotes the solution <strong>of</strong> the discretised PDE, R(uh,s) = 0, which depends upon the node positions<br />
s. The sensitivities<br />
� ∂R<br />
∂uh<br />
DJ ∂J ∂uh ∂J<br />
= +<br />
Ds ∂uh ∂s ∂s<br />
� T<br />
∗ Chemnitz University <strong>of</strong> Technology, Germany<br />
† School <strong>of</strong> Computing, University <strong>of</strong> Leeds, UK<br />
= ∂J<br />
∂s<br />
∂R<br />
− ΨT , (1)<br />
∂s<br />
Ψ = ∂J<br />
, (2)<br />
∂uh<br />
Speaker: SCHNEIDER, R. 28 <strong>BAIL</strong> <strong>2006</strong><br />
✩<br />
✪