Abaqus Nonlinear FEA: Plasticity, Large Deformation, and Solver Convergence

Abaqus Nonlinear FEA: Plasticity, Large Deformation, and Solver Convergence

Nonlinear FEA is where things get real. I remember my first nonlinear analysis — it ran for two hours and then crashed with a convergence error that I had no idea how to fix. After years of working with Abaqus/Standard, I've learned that nonlinear analysis is less about the software and more about understanding what's making your model nonlinear in the first place. Let me walk you through how I set up and troubleshoot nonlinear runs.

Nonlinearity Sources

Material Nonlinearity

• Plasticity: Permanent deformation after yield
• Creep: Time-dependent deformation under constant load
• Hyperelasticity: Large elastic deformation (rubber)
• Viscoelasticity: Rate-dependent elastic behavior

Geometric Nonlinearity

• Large deformation: Shape changes significantly affect stiffness
• Large rotation: Elements rotate significantly
• Buckling: Sudden change in deformation mode
• Contact changes: New contacts form or existing contacts separate

Boundary Nonlinearity

• Contact: Stiffness changes as bodies contact/separate
• Follower forces: Load direction changes with deformation

Setting Up Nonlinear Analysis

Step Definition

1. Module: Step
2. Create Step > Static, General (NLGEOM)
3. Set:
   • Nlgeom: ON (enables large deformation)
   • Time period: 1.0 (pseudo-time for static)
   • Incrementation:
     • Initial: 0.01 (start small)
     • Minimum: 1×10⁻⁵ (cutback limit)
     • Maximum: 0.1 (largest allowed increment)
   • Max increments: 100-1000

Material Plasticity

1. Module: Property
2. Create Material:
   • Elastic: Young's modulus (E) and Poisson's ratio (ν)
   • Plastic: Yield stress vs. plastic strain table

Plasticity Data (True Stress - True Strain)

| Plastic Strain | Yield Stress (MPa) | |----------------|-------------------| | 0.000 | 250 | | 0.005 | 280 | | 0.010 | 310 | | 0.020 | 350 | | 0.050 | 420 | | 0.100 | 480 | | 0.200 | 530 | | 0.500 | 600 |

Important: Abaqus requires true stress and true strain (not engineering). Convert:

• True stress: σtrue = σeng × (1 + εeng)
• True strain: εtrue = ln(1 + εeng)

Hardening Models

1. Isotropic hardening: Yield surface expands uniformly
   • Use for monotonic loading
   • Simple, stable
2. Kinematic hardening: Yield surface translates
   • Use for cyclic loading
   • Bauschinger effect (yield in compression after tension)
3. Combined hardening: Both expand and translate
   • Use for complex cyclic loading
   • Most realistic for low-cycle fatigue

Hyperelasticity (Rubber)

1. Create Material > Mechanical > Hyperelastic
2. Select model:
   • Mooney-Rivlin: 2-parameter (N=1), good for moderate strain
   • Ogden: N=3, best for large strain (up to 700%)
   • Yeoh: 3-parameter, good for filled rubber
   • Arruda-Boyce: Good for polymers
3. Input test data or coefficients:
   • Uniaxial test data
   • Biaxial test data
   • Planar (shear) test data
4. Evaluate > Test Data:
   • Abaqus fits the model to test data
   • Check stability (no instability in strain range)

Solver Convergence

Newton-Raphson Method

Abaqus/Standard uses Newton-Raphson iteration:

1. Apply load increment (Δλ × Ftotal)
2. Assemble tangent stiffness matrix (Kt)
3. Solve: Kt × Δu = Δλ × Ftotal - Finternal
4. Update displacement: u = u + Δu
5. Check convergence:
   • Force residual: ‖Fresidual‖ < ‖Fref‖ × RTOL
   • Displacement correction: ‖Δu‖ < ‖u‖ × DTOL
6. If converged: proceed to next increment
7. If not converged: iterate again
8. If too many iterations: cutback (reduce increment size)

Convergence Controls

1. Step > Incrementation:
   • I0: Initial increments (10-20 for smooth, 5 for difficult)
   • IR: Minimum increments (cutback limit)
   • Max increments: Total allowed
2. Step > Other > Solver Controls:
   • Equation solver: Direct (default) or Iterative
   • Residual control: RTOL = 5×10⁻³ (default)
   • Displacement control: DTOL = 5×10⁻³ (default)

Automatic Stabilization

1. For unstable problems (buckling, snap-through):
   • Step > Other > Automatic Stabilization
   • Damping factor: 2×10⁻⁴ (typical)
   • Continue damping: Yes (for post-unstable)
2. Damping adds viscous force: Fdamping = c × velocity
3. Prevents sudden instability but adds artificial energy
4. Check: Artificial energy should be < 5% of strain energy

Arc-Length (Riks) Method

1. For post-buckling analysis:
   • Step > Static, Riks
2. Set:
   • Proportional load: Load that scales with arc-length
   • Initial arc-length: 0.1 (typical)
   • Minimum arc-length: 1×10⁻⁵
3. Riks method follows the equilibrium path through limit points
4. Use for: Buckling, snap-through, snap-back

Contact in Nonlinear Analysis

General Contact

1. Module: Interaction
2. Create Interaction > Surface-to-Surface Contact (Standard)
3. Or use General Contact:
   • Interaction > General Contact
   • Automatically detects all contact pairs
4. Set:
   • Contact formulation:
     • Penalty: Default, robust
     • Augmented Lagrange: More accurate, slightly slower
     • Direct (Lagrange): Exact, but can oscillate
   • Friction: Coulomb μ (0.1 for steel-steel, 0.3 for rubber-steel)
   • Normal behavior: Hard (default) or soft (pressure-overclosure)

Contact Properties

1. Interaction Property > Contact:
   • Normal behavior:
     • Hard contact: No penetration, no tension
     • Soft contact: Pressure-overclosure relationship
     • Exponential: For soft contact
   • Tangential behavior:
     • Penalty: μ (friction coefficient)
     • Static-Kinematic Exponential Decay: μs → μk
     • Rough: No slip (bonded after contact)
   • Damping: Contact damping for stabilization

Contact Troubleshooting

1. Chattering: Contact opens and closes repeatedly
   • Fix: Add contact damping, use penalty formulation, refine mesh
2. Penetration: Slave nodes penetrate master surface
   • Fix: Increase penalty stiffness, use augmented Lagrange
3. No convergence in contact: Too many contact changes per increment
   • Fix: Reduce increment size, add stabilization, refine mesh at contact

Output and Post-Processing

Field Output

1. Field Output Requests:
   • Stresses: S (stress components), SINV (von Mises, principal)
   • Strains: E (total), PE (plastic), EE (elastic)
   • Displacement: U
   • Contact: CPRESS, CSHEAR, COPEN, CSLIP
2. Save at:
   • Every increment: For detailed history
   • Last increment only: For final state (saves space)

History Output

1. History Output Requests:
   • Reaction force: RF at constrained nodes
   • Displacement: U at specific nodes
   • Energy: ALLKE, ALLIE, ALLPD, ALLVD, ALLWK
   • Contact: CPRESS at specific contact nodes
2. History output is per-step, not per-element (lighter on storage)

Energy Balance Check

1. For nonlinear analysis, verify energy balance:
   • External work (ALLWK) ≈ Internal energy (ALLIE) + Plastic dissipation (ALLPD) + Viscous dissipation (ALLVD)
2. If artificial energy (ALLVD) > 5% of ALLIE:
   • Stabilization is too high
   • Reduce damping factor
   • Results may be inaccurate

Convergence Troubleshooting

Common Convergence Issues

Issue: Too Many Cutbacks

Symptom: Increment size keeps reducing, analysis takes very long. Fix:

1. Check contact (chattering, penetration)
2. Add automatic stabilization (damping)
3. Reduce initial increment size
4. Refine mesh in high-strain regions
5. Use mass scaling (for quasi-static)

Issue: No Convergence at Specific Load

Symptom: Analysis converges until a certain load, then fails. Fix:

1. Check for material instability (negative tangent modulus)
2. Check for buckling (use Riks method)
3. Check for contact separation (add stabilization)
4. Check for excessive element distortion (remesh)

Issue: Excessive Element Distortion

Symptom: Elements become too distorted, solver fails. Fix:

1. Refine mesh in high-deformation regions
2. Use reduced integration elements (C3D8R instead of C3D8)
3. Enable adaptive remeshing (Abaqus/Explicit)
4. Use ALE (Arbitrary Lagrangian-Eulerian) for large deformation

Element Selection for Nonlinear Analysis

| Element | Type | Use | |---------|------|-----| | C3D8R | Hex, reduced integration | General 3D (default for nonlinear) | | C3D8 | Hex, full integration | When hourglass is problematic | | C3D8H | Hex, hybrid | Rubber, incompressible materials | | C3D20R | Hex, quadratic, reduced | High accuracy (stress concentration) | | C3D10M | Tet, modified | Complex geometry, good for contact | | S4R | Shell, reduced | Thin walls, sheet metal | | B31 | Beam, linear | Frame structures |

Hourglass control: C3D8R elements can exhibit hourglass modes (zero-energy deformation). Check artificial hourglass energy < 5% of internal energy.

Verification Checklist

• [ ] Nlgeom is ON for large deformation
• [ ] Material plasticity uses true stress-strain data
• [ ] Hardening model matches loading type (monotonic vs. cyclic)
• [ ] Contact formulation is appropriate (penalty for general, augmented for accuracy)
• [ ] Initial increment is small (0.01-0.05)
• [ ] Automatic stabilization is used for unstable problems
• [ ] Artificial energy < 5% of internal energy
• [ ] Hourglass energy < 5% of internal energy (reduced integration)
• [ | Reaction forces balance applied loads
• [ ] No excessive element distortion
• [ ] Energy balance is satisfied

Wrapping Up

If I had to boil nonlinear FEA down to one thing, it's this: know what's making your model nonlinear and set up your solver accordingly. Small initial increments, reasonable cutback limits, and stabilization when you need it. I check the energy balance on every nonlinear run — if artificial energy is more than 5% of internal energy, something's off with your damping or your mesh. And don't be afraid to start with a simpler model and add complexity. I've seen too many people throw a fully nonlinear model at the solver on day one and then spend a week trying to figure out why it won't converge.