# FIELD CONSISTENCY

-M.SURENDRAN (QHS 016)

• 3D problems  2D problems
– Plane stress (dimension small in normal direction) – Plane strain (one dimension is very long) – Axisymmetric (Involving solids of revolution)

• Sufficient accuracy
– Linear 3-noded triangle – Bilinear 4-noded quadrilateral

• Higher accuracy by higher order elements

Problems with current approach • Plane stress modelling of beam flexure • Poisson’s ratio ≈ 0.5 (nearly incompressible materials and materials undergoing plastic deformation) • Parasitic shear – Similar to shear locking in 1D Timoshenko beam element – Error increases with increase in aspect ratio • Solved by Reduced integration • Addition of bubble modes for 4-noded rectangular elements .

• 8-noded and 9-noded elements – No locking with reduced integration – Show severe linear and quadratic shear stress oscillations • Shear locking. parasitic shear  field inconsistency • Adding drilling degrees of freedom (rotational DOF normal to the plane of element) • Why does the technique work? . membrane locking.

poor convergence and violent stress oscillations in c0 elements formulation are due to lack of a consistent definition of critical strain field (strain field constrained in penalty regime) • Eg. Shear strain in Timoshenko beam and Mindlin plates and shells. volumetric strain in incompressible elasticity . membrane strain in curved beams and shells.Problems due to inconsistency • Locking.

• The error in a timoshenko beam increases as the beam becomes thin even if shape functions are chosen to satisfy completeness and continuity • As the beam becomes thin the shear strains should vanish and automatically enforce kirchhoff’s codition .

Timoshenko beam  Shear stiffness becomes large as depth decreases • So high rank and non-singularity • But these are the outcomes of assumptions in formulation.Symptoms vs Causes • Locking  FEM solution vanishes quickly with increase in penalty multipliers • Locking is linked to the non-singularity of stiffness matrix • Eg. Unexpected requirements are unsatisfied (symptoms) • Not the reason that fem does not • Cause??? .

FIELD CONSISTENCY • So the penalty linked strain fields should be discretized in a consistent way • Only physically meaningful constraints • “The requirement that a certain strain field interpolation may have to be defined in a manner that only physically realistic constraint conditions will emerge in constrained physical regimes” .

BEAM THEORY • • • • • • Transverse deflection of thin cantilever beam Length = L Load = q N/m A A A .

By Ritz approximation • • • • • A A A This is ‘best-fit’ or ‘least-squares fit’ This seeks the best approximation .

THE TIMOSHENKO BEAM THEORY • General formulation of beam flexure with transverse shear deformation • Total strain energy  2 independent functions  w(x) and θ(x) • A • A .

introduces a penalty condition on the shear strain] • a10 imposes zero bending strain [spurious] . a1. b1  0 [ε  ∞.2-TERM RITZ APPROXIMATION (Inconsistent) • • • • • • θapprox = a1x u = b1x γapprox = a1x – b1 a1= -3qL2/(12EI + εL2) b1= -qL/2ε – 1.5qL3/(12EI + εL2) As ε ∞.

(a1 – 2b2)  0. No spurious constraint on bending strain (θapprox ) • BM is constant approximation of the quadratic variation • SF has linear variation even if ε  ∞ .3-TERM RITZ APPROXIMATION (consistent) • • • • θapprox = a1x u = b1x + b2x2 γapprox = -b1 + (a1 – 2b2)x So if shear strain  0 .

x) So constant BM Same as 3-term approximation Quadratic shear oscillations. (which correspond to 2 point Gaussian integration rule) .4-TERM RITZ APPROXIMATION (Inconsistent) • • • • • • • • θapprox = a1x + a2x2 u = b1x + b2x2 γapprox = -b1 + (a1 – 2b2)x + a2x2 ε  ∞. But they vanish at ξ = ± 1/√3. shear strain  0 . a2  0 This means constant bending strain(θapprox .

a2L2/6) + (a1 + a2L – 2b2)x Yield physically meaningful constraints But we see that approx solution should be judged correctly .4-TERM RITZ APPROXIMATION (Consistent) • • • • • • Assume such that shear strain is consistent Inconsistent term is quadratic term in a2 Replace x2 by (Lx – L2/6) γapprox = (-b1 .

• Explained easily by shear flexible beam element • Using reduced integration produced accurate elements • It introduced errors that compensate the other constraining errors • Functional reconstitution  procedure to derive errors resulting from use of inconsistent strain field interpolations .

u 2l L X.v Strain energy functional consists of energy from normal strain (UE) and shear strain (UG)` .U 2t T • • • • • Length = L Depth = T Thickness = t DOF = u.PLANE STRESS MODEL OF BEAM FLEXURE Y.V y.v x.

shear force is finite • This is the condition that causes difficulties .• a • Timoshenko beam theory is obtained by substituting u = yθ • For slender beams shear strains=0 (Classical Euler – Bernoulli beam theory) • But shear stress.

x = (a2 +b1) + a3x + b3y .y + v. v1 – v4 • Field variables are interpolated as • u = a0 + a1x + a2y + a3xy • v = b0 + b1x + b2y + b3xy • Problem is associated with shear strain as it will be constrained and will become vanishingly small when flexural action of thin beam is modelled • γ = u.4-NODED RECTANGULAR ELEMENT • 4-nodes • 8-DOF  u1 – u4.

• Shear strain energy within element will be • E • Since shear strain = 0 in thin beam • a2 + b1  0 • a3  0 • b3  0 • First constraint imposes conditions on both u and v • Second and third impose conditions separately • These are undesirable  stiffening effect  parasitic shear .

then a3  0 will be enforced more rapidly than b3  0 • The spurious energy generated from these terms will be in similar proportions • Since shear force at a section along length is constant shear strain will also be constant .• For slender beams (l > t)  l2 >>t2 • If shear strain energy approaches zero in thin beams.

• EQ • Constant term  averaged shear strain • Linear oscillating termrelated to the spurious constraint • This oscillation is self-equilibriating over the element. • But it contributes a finite energy in equation • In slender beams it becomes major and dominates behavior of beam .

• e= e is ratio of actual a3 and consistent a3 .• The discretized strain energy functional of a beam portion from to can be re-constituted as • A very simple ‘trick’ is to use a one point integration rule for the shear strain energy.

• By comparing the displacement eqns with actual displacements we get a3.consis=M0/EI • average shear-stress representation in a 4node field-inconsistent plane stress element as • For long element (slender beam) • It also produces shear oscillations .

8-NODED RECTANGULAR ELEMENT • Quadratic shear stress oscillations appear • The performance of this element can be improved by reduced integration of the shear strain energy • A • a • γ= .

• Oscillating term is • If a6  0 it will create inconsistency • the quadratic function is an oscillating function with zero-points at the Gauss points associated with the 2-point integration rule • FET  Piece wise Ritz method stresses taken at gauss points  so accurate .