Universität ität Stuttgarttt t Fakultät Bau und- Umweltingenieurwissenschaften Baustatik und Baudynamik On the Sensitivity of Finite Elements to Mesh Distortions Manfred Bischoff, Benjamin Schneider Deepak Ganapayya, Nagesh Chilakunda Institute of Structural Mechanics University of Stuttgart bischoff@ibb.uni-stuttgart.de i tt t 1
Universität ität Stuttgarttt t Fakultät Bau und- Umweltingenieurwissenschaften Baustatik und Baudynamik motivation locking phenomena in structural and solid finite elements optimal technologies available for rectangular element shapes different technologies are identical for rectangular element shapes sensitivity to mesh distortions is one of the last open problems (further challenges in non-linear analysis and 3d-shells) Motivation 2
How to Deal with this Problem 1. avoid mesh distortions complex geometries adaptive (re-) meshing Motivation 3
How to Deal with this Problem 1. avoid mesh distortions complex geometries adaptive (re-) meshing 2. develop finite elements which are insensitive to mesh distortions two-dimensional elasticity four-node elements Motivation 4
Measuring Distortion Sensitivity the most popular test: bending of a cantilever insensitive sensitive various versions in the literature Motivation 5
Element Types under Investigation four-node plane stress elements Q1 standard Galerkin formulation suffers from shear locking and volumetric locking Q1-SRI selective reduced integration of shear part no shear locking, o.k. for small Poisson s ratio Qm6 method of incompatible modes ( = Q1-E4) locking-free (?) Taylor, Wilson (1973, 1976), Simo et al. (199) Motivation 6
Two Element Bending Test typical results beam theory Qm6 Q1-SRI Q1 Measuring Distortion Sensitivity 7
Two Element Bending Test how meaningful is this test setup? is this a distorted mesh? stresses in quadrilaterals are identically zero! undistorted elements different results Measuring Distortion Sensitivity 8
Two Element Bending Test how meaningful is this test setup? numerical experiment is restricted to specific loads and boundary conditions constant-linear stress distribution principal stresses aligned to edges exact solution = beam solution may be a hint toward optimization of element technology but not a guideline Measuring Distortion Sensitivity 9
Element Quality Tendency to Locking eigenvalues as objective measure of element quality (no sum on i) eigenvalue = stiffness eigenvector D = deformation mode no need to choose specific loads and boundary conditions eigenvalue spectrum = element deformation spectrum Measuring Element Quality 1
Eigenvalue Analysis shear locking trapezoidal mode optimal: volumetric locking optimal: only one eigenvalue with distortion sensitivity what are the correct values of for arbitrary element shapes? Measuring Element Quality 11
Considering a Patch of Finite Elements 9 elements, 16 nodes, 32 d.o.f. spectrum of 32 eigenvalues undistorted, optimal mesh provides reference solution distorted mesh perturbed eigenvalue spectrum Distortion Patch Test 12
Distortion Patch Test removing internal d.o.f. via static condensation static condensation of node 13 16 macro element with Distortion Patch Test 13
Distortion Patch Test properties of macro element 24 eigenpairs depend on locations of invisible nodes 13 16 comparison of and yields objective measure for distortion sensitivity Distortion Patch Test 14
Element Distortion 8 generic distortion modes = 8 single element modes Distortion Patch Test 15
Application of Distortion Patch Test numerical experiments with Q1, Q1-SRI and Qm6 trapezoidal type of distortion plotting versus comparing eigenvalue spectra for different values of Application 16
Eigenvalues versus Distortion Parameter standard Galerkin finite element Q1 maximum distortion 18 16 14 12 1 8 6 4 2.1.2.3.4.5.6 mode 1 mode 2 mode 3 mode 4 mode 5 mode 6 mode 7 mode 8 mode 9 mode 1 mode 11 mode 12 mode 13 mode 14 mode 15 mode 16 mode 17 mode 18 mode 19 mode 2 mode 21 Element Type Q1 17
Eigenvalues versus Distortion Parameter modes 1 4 4 mode 1, 3 2 mode 1 1 mode 2 mode 3 mode 4.1.2.3.4.5.6 Element Type Q1 18
Eigenvalue Spectra standard Galerkin finite element Q1 18 16 14.1.2 12.3 1.4.5 8.6 6 4 2 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 19 2 21 mode Element Type Q1 19
Application to Thin Structure more sensitive to locking 3 25 2 15 mode 1 mode 2 mode 3 mode 4 mode 5 mode 6 mode 7 mode 8 mode 9 mode 1 mode 11 mode 12 mode 13 1 mode 14 5 4 8 12 mode 15 mode 16 mode 17 mode 18 mode 19 mode 2 mode 21 Element Type Q1 2
Application to Thin Structure modes 1 4 1 8 6 mode 1 mode 2 mode 3 mode 4 mode 5 4 2 5 1 Element Type Q1 21
Eigenvalue Spectrum of Thin Structure standard Galerkin finite element Q1 3 25 2 15 1 2 4 6 8 1 12 5 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 19 2 21 mode Element Type Q1 22
Eigenvalue Spectrum of Thin Structure standard Galerkin finite element Q1 15 1 2 4 6 8 1 12 5 1 2 3 4 5 6 7 mode Element Type Q1 23
Eigenvalue Spectrum of Thick Structure for comparison 7 6 5 4 3 1.1.2.3.4.5.6 2 1 1 2 3 4 5 6 7 mode Element Type Q1 24
Selective Reduced Integration, Q1-SRI eigenvalues for thick structure 18 16 14 12 1 8 6 4 2.1.2.3.4.5.6 mode 1 mode 2 mode 3 mode 4 mode 5 mode 6 mode 7 mode 8 mode 9 mode 1 mode 11 mode 12 mode 13 mode 14 mode 15 mode 16 mode 17 mode 18 mode 19 mode 2 mode 21 Element Type Q1-SRI 25
Selective Reduced Integration, Q1-SRI eigenvalues for thick structure 45 4 35 3 25 2 15 1 5 mode 1 mode 2 mode 3 mode 4 mode 5.1.2.3.4.5.6 Element Type Q1-SRI 26
Selective Reduced Integration, Q1-SRI eigenvalues for thin structure 3 25 2 15 5 5 1 mode 1 mode 2 mode 3 mode 4 mode 5 mode 6 mode 7 mode 8 mode 9 mode 1 mode 11 mode 12 mode 13 mode 14 1 mode 15 mode 16 mode 17 mode 18 mode 19 mode 2 mode 21 Element Type Q1-SRI 27
Selective Reduced Integration, Q1-SRI eigenvalues for thin structure 3 2 mode 1 mode 2 mode 3 mode 4 mode 5 mode 6 mode 7 1 5 1 Element Type Q1-SRI 28
Selective Reduced Integration, Q1-SRI eigenvalue spectrum 3 25 2 15 2 4 6 8 1 12 1 5 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 19 2 21 mode Element Type Q1-SRI 29
Selective Reduced Integration, Q1-SRI eigenvalue spectrum 15 1 2 4 6 8 1 12 5 1 2 3 4 5 6 7 mode Element Type Q1-SRI 3
Method of Incompatible Modes, Qm6 (=Q1-E4) eigenvalues for thin structure 25 2 15 1 5 1 2 3 4 5 6 7 8 9 1 11 12 mode 1 mode 2 mode 3 mode 4 mode 5 mode 6 mode 7 mode 8 mode 9 mode 1 mode 11 mode 12 mode 13 mode 14 mode 15 mode 16 mode 17 mode 18 mode 19 mode 2 mode 21 Element Type Qm6 31
Method of Incompatible Modes, Qm6 (=Q1-E4) eigenvalues for thin structure 3 mode 1 mode 2 mode 3 mode 4 2 mode 5 mode 6 mode 7 1 1 2 3 4 5 6 7 8 9 1 11 12 Element Type Qm6 32
Method of Incompatible Modes, Qm6 (=Q1-E4) eigenvalues for thin structure 25 2 15 1 2 4 6 8 1 12 5 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 19 2 21 mode Element Type Qm6 33
Method of Incompatible Modes, Qm6 (=Q1-E4) eigenvalues for thin structure 15 1 2 4 6 8 1 12 5 1 2 3 4 5 6 7 mode Element Type Qm6 34
Summary of Results eigenvalue spectra 3 3 Q1 Q1-SRI Qm6 3 25 25 25 2 2 2 15 15 15 1 1 1 5 5 5 1 6 11 16 21 mode 1 6 11 16 21 mode 1 6 11 16 21 mode Summary and Conclusions 35
Summary of Results eigenvalue spectra 15 Q1 Q1-SRI Qm6 15 15 1 1 1 5 5 5 1 2 3 4 5 6 7 mode 1 2 3 4 5 6 7 mode 1 2 3 4 5 6 7 mode Summary and Conclusions 36
Summary of Results eigenvalues of mode 1 in dependence of mesh distortion d 3 25 2 15 Q1 Q1-SRI 1 5 Qm6 5 1 Summary and Conclusions 37
Summary of Results eigenvalues of mode 1 in dependence of mesh distortion d 3. 2. Q1 1. Q1-SRI Qm6. 1 2 3 Summary and Conclusions 38
Summary of Results eigenvalues of mode, thick structure (square) 3 25 2 Q1 Q1-SRI 15 Qm6 1 5.1.2.3.4.5.6 Summary and Conclusions 39
Conclusions measuring distortion sensitivity bending test is not so bad eigenvalues may provide objective measure newly proposed distortion patch test objective measure for distortion sensitivity universal locking test applicable to arbitrary problems (thin and curved structures, near incompressibility, etc.) numerical results distortion sensitivity is related to locking Q1 and Q1-SRI equally sensitive to distortion Qm6 significantly better than Q1-SRI (more than bending test implies) Summary and Conclusions 4