Presets
Playback
Parameters
Camera
Layers
Quality
Model check
The science
What is modelled
A 4 m long, 1 m deep, 50 mm thick aluminium-like plate is clamped at its left edge and loaded in its own plane at the free end (plane stress). The finite element method cuts the plate into small elements, writes the elastic energy of each one as a stiffness matrix, assembles them into one global system K u = F and solves it for the displacement of every node. You can watch a conjugate-gradient solver build the answer iteration by iteration, compare it with the exact direct solution, and refine the mesh to see the answer approach beam theory.
Equations
- Plane-stress law:
σ = D ε,D = E/(1−ν²) [[1, ν, 0], [ν, 1, 0], [0, 0, (1−ν)/2]]; strain from displacementε = B ue. - Triangle (CST) element:
Ke = t A BT D BwithBconstant over the element. Quadrilateral (Q4): bilinear shape functions,Ke = ∫ BT D B t dAby 2 × 2 Gauss integration. - Assembly: the element matrices are added into the global matrix at the degrees of freedom of their nodes; the clamped edge sets
u = v = 0. Nodes are numbered along the depth, so the matrix is banded (see the last graph). - Loads: a tip load P is shared over the free edge as consistent nodal forces; the moment case applies the linear edge traction
σx = M y / IwithM = P L. - Solvers: banded Cholesky factorisation (the exact FE answer) and the conjugate-gradient iteration
α = r·r / p·Kp, u += α p, r −= α Kp, β = rnew·rnew / r·r, p = r + β p, stopped at a relative residual of 1e-10. - Beam theory reference,
I = t h³/12: tip loadδ = P L³/(3 E I) + P L/(κ G A)(bending plus shear, κ = 5/6,G = E/(2(1+ν))); end momentδ = M L²/(2 E I); root bending stressσ = M (h/2)/I. - Von Mises stress for the colours:
σvM = √(σx² − σxσy + σy² + 3τ²)at the element centre.
Checks run in the page
- Patch test: a linear displacement field imposed on the boundary must be reproduced exactly inside, for both element types.
- Equilibrium: clamp reactions balance the load in force and in moment.
- Energy: external work equals the strain energy.
- Refinement: the tip deflection approaches the beam-theory value monotonically on four meshes.
- Conjugate gradients reach the direct solution, and doubling E halves the deflection.
Honest simplifications
- Linear elasticity and small displacements only; the picture exaggerates the displacement by the factor on the slider (default 150), the real tip moves a few millimetres.
- Coarse triangles are far too stiff in bending (shear locking: 63 % of the beam-theory deflection on 8 × 2 elements); that is a real property of the element, not a bug, and the refinement graph shows how slowly it goes away.
- Beam theory is itself an approximation for a plate this deep (h/L = 1/4) with a fully clamped edge, so even a perfect FE solution differs from it by a few per cent; the page reports the ratio, not an error against truth.
- Stresses are element-centre values and the peak at the clamp is under-resolved on coarse meshes. Real clamps also have stress concentrations that this model does not capture. This is not a design tool.
What to try
- Start with the coarse preset and refine: watch the tip deflection climb towards beam theory.
- Switch to quadrilaterals at the same mesh and see how much closer they are.
- Use the end-moment case: pure bending is a clean test where the answer is known exactly.
- Raise the Poisson ratio and watch the shear part of the deflection change.
Loading model...
Loading Blender model...