A nonlinear material model tells a finite element analysis how the metal behaves once it yields. The usual choices are elastic-perfectly-plastic, used for limit-load checks because it ignores strain hardening and so gives a cautious collapse load, and hardening curves, bilinear or true stress-strain, used where how far the metal stretches matters, such as elastic-plastic collapse and local strain checks. A yield criterion such as von Mises turns a multiaxial stress into one equivalent value, and a hardening rule says where yield sits when the load reverses.
This article goes through each choice, what it is for and what it can hide, and ends with geometric nonlinearity, the other switch that collapse and buckling checks usually need.
Elastic-perfectly-plastic
The simplest model is a straight elastic line up to yield, then flat: strain can grow with no further rise in stress. It needs only two numbers, the elastic modulus and a yield value, and it is robust in a solver.
Because it ignores strain hardening, the collapse load it gives is on the cautious side for steels that harden. Limit-load checks use it, with the yield value set by the governing code; that value is often a multiple of the allowable stress rather than the measured yield, so the margin is built into the material curve.
Hardening curves: bilinear and true stress-strain
Real steel gets stronger as it is stretched past yield. A bilinear model draws this as a second, shallower straight line, defined by a tangent modulus. A true stress-strain curve follows the measured shape, or a code's standard formula for it, smoothly bending over as strain grows.
Test data usually come as engineering stress and strain, based on the original cross-section and length. The solver needs true values, based on the current, shrinking section. Up to the start of necking, while the strain is spread evenly along the specimen, they convert as:
Hardening curves are used for elastic-plastic collapse and local strain checks, where how far the metal stretches matters as much as when it first yields.
One number from many stresses
At any point the metal carries normal stresses in three directions as well as shears. A yield criterion turns them into one equivalent stress to compare with the single yield value from a tensile test. The von Mises criterion, based on the energy that changes the metal's shape, is the usual choice. Tresca, based on the largest shear stress, is slightly more cautious: for the same stresses it never gives a lower equivalent stress, and at most about 15 % higher.
When loads cycle: hardening rules
Once the metal has yielded, the model must know where yield sits when the load reverses. Isotropic hardening grows the yield range evenly in both directions, so after stretching to a higher stress the metal must be pushed equally far the other way before it yields again. Kinematic hardening shifts the yield range instead, so stretching past yield lowers the yield in reverse, the Bauschinger effect seen in real steels. Combined models mix the two.
For cycling loads and ratcheting checks, kinematic hardening is closer to how steel behaves. Isotropic hardening can make later cycles look elastic and hide ratcheting, so it should be used for cyclic checks only where the governing code allows it. Codes may prescribe the model for these checks, in some cases elastic-perfectly-plastic.
Geometry is another nonlinearity
Material nonlinearity is about the metal; geometric nonlinearity is about the shape. When deflections grow large, the loads act on a changed shape. A flat plate under pressure stiffens as it bulges and starts to carry the load as a membrane, while a compressed shell softens and can buckle. Pressure that follows the deforming surface is part of the same effect.
The two are separate switches in the model. Collapse and buckling checks usually need both on: large displacements, so that the shape change is captured, and a yielding material, so that plasticity is.
Matching the model to the check
| Check | Material model | Geometry |
|---|---|---|
| Limit load | Elastic-perfectly-plastic, code yield value | Small displacement |
| Elastic-plastic collapse | True stress-strain with hardening | Large displacement |
| Local strain | True stress-strain with hardening | Large displacement |
| Ratcheting | Perfectly plastic or kinematic hardening | As the code directs |
| Buckling with imperfections | Yielding material, often with hardening | Large displacement |
Key points
- Elastic-perfectly-plastic is simple and cautious, and suits limit-load checks.
- Hardening curves, bilinear or true stress-strain, suit collapse and local strain checks; convert test data to true values.
- von Mises is the usual yield criterion; Tresca is slightly more cautious.
- Kinematic hardening is closer to steel under cycling; isotropic hardening can hide ratcheting.
- Geometric nonlinearity is a separate switch; collapse and buckling checks usually need it on.
Where this matters in practice
- Limit-load checks of nozzles, openings and junctions in heads and shells.
- Elastic-plastic collapse and local strain checks at heavily loaded lugs, supports and attachments.
- Buckling checks of thin shells, cones and heads with an imposed imperfection.
- Ratcheting checks at attachments and junctions under cycling loads.
Corix includes a limit-load analysis on a shell model, using the elastic-perfectly-plastic model with small displacements. In early access, by invitation.
See the specialty moduleRelated principles
- Elastic and elastic-plastic analysisTwo ways to judge a computed stress: sort it against limits, or let the metal yield.5 min read
- Linear eigenvalue buckling analysisHow a computer model finds a buckling load multiplier, and why its answer needs reducing.5 min read
- Ratcheting and shakedownWhat happens when a steady stress meets one that cycles: shakedown, plastic cycling or ratcheting.6 min read
Next principle: Local loads on pressure vessel shells
Educational article. It explains principles and is not design guidance; follow your governing code and edition.