Quick Answer
Non-linear analysis failures in ABAQUS and SAP2000 fall into four categories: convergence divergence, contact instability, material instability, and numerical singularity. Fix convergence in ABAQUS by reducing the load increment to 1–5% of total load and enabling *CONTROLS, PARAMETERS=FIELD with tightened tolerances. In SAP2000, reduce the displacement increment to ≤0.001 of the target and verify hinge property completeness. Both platforms require correct P-Delta flags, adequate mesh density (≥4 elements per plastic hinge zone), and material data covering the full strain range.
Non-linear finite element analysis is where structural engineering theory meets brutal computational reality. Whether you are running a seismic pushover on a reinforced-concrete frame in SAP2000 or a ductile fracture simulation in ABAQUS/Standard, the dreaded “analysis did not converge” message can cost hours — time most project schedules cannot afford.
This guide maps every common failure mode to a specific, actionable fix. It covers the underlying numerical mechanics so you understand why a solution diverges — not just what button to press. Use it alongside the SAP2000 Pushover Analysis guide and the ABAQUS Plasticity Models comparison on this site.
1. Why Non-Linear Analysis Fails: The Numerical Mechanics
All finite element solvers iterate toward equilibrium. In a linear analysis, the stiffness matrix [K] is constant — one matrix factorization solves the system. In a non-linear analysis, [K] changes with every deformation increment, and the solver must repeatedly update and re-invert it. The most widely used algorithm is Newton-Raphson (NR) iteration — at each load increment it computes the residual (difference between applied and internal forces) and checks it against a tolerance.
When the solver fails, it is telling you one of three things: the tangent stiffness matrix is singular (a mechanism has formed), the load step is too large, or the material/contact model is returning physically inadmissible states (negative tangent stiffness from softening or snap-through).
2. Types of Non-Linearity: A Structural Engineer’s Map
| Type | Physical Source | ABAQUS Implementation | SAP2000 Implementation | Cost |
|---|---|---|---|---|
| Geometric (GN) | Large deformations, P-delta, buckling | *STEP, NLGEOM=YES | P-Delta load case / Large displacements flag | Moderate |
| Material (MN) | Yielding, cracking, creep, damage | *PLASTIC, *DAMAGE cards | Nonlinear hinge assignments (FEMA 356 / user-defined) | High |
| Contact (CN) | Opening/closing interfaces, friction | *CONTACT PAIR, General Contact | Gap/link elements, frame releases | Very High |
| Boundary (BN) | Changing supports (rocking foundations) | Amplitude-driven BC changes | Staged construction, nonlinear supports | Moderate |
| Combined GN+MN | Post-yield large deformation (collapse) | Both flags active simultaneously | Pushover with P-Delta | Extreme |
3. ABAQUS Convergence Failures — Diagnosis & Fixes
ABAQUS/Standard uses a full Newton-Raphson scheme with automatic increment control. The critical output file is the .msg file — not the .odb. Open it immediately when convergence fails; it lists the node and degree of freedom with the largest residual.
3.1 Load Increment Strategy
The most common error is too-large initial increments. ABAQUS default initial increment is 1.0 (the full load in one step). For any moderately nonlinear problem this will diverge.
| Problem Type | Initial Inc. | Min Inc. | Max Inc. | Max INC |
|---|---|---|---|---|
| Mild MN (elastic-perfectly plastic) | 0.05 | 1×10⁻⁵ | 0.1 | 200 |
| Moderate MN (combined hardening) | 0.01 | 1×10⁻⁶ | 0.05 | 500 |
| Severe MN / post-peak softening | 0.001 | 1×10⁻⁸ | 0.02 | 2000 |
| Contact-dominated | 0.005 | 1×10⁻⁷ | 0.02 | 1000 |
| Dynamic implicit (HHT) | Δt/T ≤ 0.01 | 1×10⁻⁸ | Δt/5 | 200 |
3.2 Contact Problems
Contact is the leading cause of convergence failures in ABAQUS. The solver must detect and enforce zero-penetration constraints at every iteration, and contact state (open/closed) can oscillate between iterations.
COPEN oscillates between positive and negative values.- Switch from hard contact to linear pressure-overclosure for the first debug run: *SURFACE BEHAVIOR, PRESSURE-OVERCLOSURE=LINEAR
- Add *CONTACT CONTROLS, AUTOMATIC TOLERANCES to let ABAQUS self-tune contact tolerances.
- Ensure the slave surface mesh is finer than the master surface (≤50% of master element size).
- For friction μ > 0.4, use *FRICTION, ROUGH only after full normal contact is established in an earlier step.
3.3 Material Instability — Concrete Damage Plasticity
| CDP Parameter | Symbol | Typical Range | Default | Impact if Wrong |
|---|---|---|---|---|
| Dilation angle | ψ | 25°–45° | 30° | Excessive volumetric expansion → divergence |
| Eccentricity | ε | 0.05–0.2 | 0.1 | Unstable flow rule near hydrostatic axis |
| Compressive/tensile ratio | f_b0/f_c0 | 1.10–1.16 | 1.16 | Wrong biaxial yield locus |
| K (yield surface shape) | K_c | 0.64–0.80 | 0.667 | Ill-conditioning if < 0.5 |
| Viscosity parameter | μ | 0–0.0002 | 0 | μ > 0.001 over-regularizes, masks damage |
3.4 Automatic Stabilization
4. SAP2000 Convergence Failures — Diagnosis & Fixes
4.1 Pushover Analysis Errors
| Error / Symptom | Root Cause | Fix |
|---|---|---|
| “Analysis failed to converge at step N” | Displacement increment too large relative to hinge softening slope | Reduce max displacement increment to 0.1–0.5 mm; increase max steps to 2000 |
| Capacity curve stops before target displacement | Mechanism formed (correct) OR numerical locking (incorrect) | Check Load Application Control — ensure displacement-controlled. Enable P-Delta. |
| Negative stiffness at step 1 | P-Delta gravity preload not in equilibrium | Run gravity case as nonlinear static first; use as initial condition for pushover |
| Capacity curve flat from the start | No hinges on lateral-load-resisting members | Assign auto-hinge (FEMA 356 Table 5-6 for beams, 5-8 for columns) or PMM hinges |
| Oscillating base shear | Multiple hinges with identical deformation capacity forming simultaneously | Offset hinge acceptance criteria by ±5% to avoid simultaneous formation |
4.2 P-Delta vs Large Displacements
| Feature | P-Delta | P-Delta + Large Displacements |
|---|---|---|
| Column axial load × lateral drift | ✓ | ✓ |
| Beam axial load × chord rotation | ✗ | ✓ |
| Accurate for drift < H/50 | ✓ | ✓ |
| Accurate for drift > H/50 | ✗ | ✓ |
| Required per ASCE 41 | ✓ minimum | ✓ preferred |
| When to use | Regular frames, drift < H/50 | Tall/irregular frames, post-peak |
4.3 Plastic Hinge Definition Issues
- Wrong hinge type: Use PMM hinges for columns with N/N_max > 0.2 — M3-only misses axial-moment interaction.
- Residual strength C-D slope: Must be defined; leaving it at 0 causes the curve to spike then collapse numerically.
- Strain hardening (B-C slope): Set to 3–5% of elastic stiffness. Zero creates a numerically ambiguous perfectly plastic hinge.
5. ABAQUS vs SAP2000: Non-Linear Solver Comparison
| Criterion | ABAQUS/Standard | SAP2000 | Winner |
|---|---|---|---|
| Solver algorithm | Full Newton-Raphson + quasi-NR option | Modified Newton-Raphson | ABAQUS (quadratic convergence) |
| Auto increment control | ✓ Very robust | ✓ Good | ABAQUS |
| Contact modeling | ✓ General contact, hard/soft | ⚠ Gap elements only | ABAQUS |
| Material model library | CDP, Drucker-Prager, VUMAT, Mohr-Coulomb, hyperelastic, creep | Kinematic hardening, fiber models, user hinge | ABAQUS |
| Pushover (seismic) | ⚠ Manual setup | ✓ Native FEMA 356 / ATC-40 | SAP2000 |
| P-Delta integration | NLGEOM=YES | Dedicated P-Delta load case | SAP2000 (easier workflow) |
| Debugging output | .msg file — extremely detailed | Analysis log — moderate detail | ABAQUS |
| Learning curve | Steep | Moderate | SAP2000 |
| Typical use | Research, component-level, advanced materials | Building-level seismic performance, design office | Context-dependent |
6. Key Convergence Criteria & Formulas
7. Pre-Run Diagnostic Checklist
- Mesh quality: aspect ratio < 10:1 for hex elements; < 5:1 near contact zones
- Element type: C3D8R for plasticity; C3D8H (hybrid) for near-incompressible materials
- Material data covers full strain range to fracture — no extrapolation beyond last point
- No over-constrained nodes (duplicate BC in same DOF)
- Contact: slave surface mesh ≤ 50% of master element size
- Load applied via smooth amplitude curve — avoid step application
- NLGEOM=YES confirmed in *STEP for large deformation cases
- Initial increment ≤ 5% of total load; minimum increment ≤ 1×10⁻⁶
- Output frequency: every 5–10 increments (not every increment)
- Run data check first: abaqus job=filename datacheck
- Gravity nonlinear case completed and locked as starting condition for pushover
- P-Delta enabled in nonlinear case parameters
- All beam and column hinges assigned at both ends (relative positions 0.0 and 1.0)
- Column hinges are PMM type for P/P_cap > 0.2
- Hinge backbone covers A through E (all five points including C-D negative slope)
- Maximum displacement increment ≤ 0.5% of target displacement
- Number of output steps ≥ 500
- Lateral load pattern matches first mode shape
- Mass source defined: from loads, with live load factor per ASCE 7
- Diaphragm assignment consistent with floor plate assumptions
8. Step-by-Step Troubleshooting Workflow
- Read the error output file first. ABAQUS: open .msg → search “NOT CONVERGED” — note increment number, iteration count, node/DOF with maximum residual. SAP2000: Analysis → Show Analysis Log.
- Locate the problem region. In ABAQUS, plot field output variable COORD at the last converged increment. In SAP2000, check the plastic hinge formation sequence.
- Reduce the load increment by 10×. If original was 0.01, set to 0.001. If this fixes it, the problem is increment size, not model error. Gradually increase back.
- Disable one nonlinearity type. Switch to elastic material (keep geometric NL), or disable contact, or remove P-Delta. Find which switch fixes it — that isolates the culprit.
- Check mesh density in the failure zone. Add at least 4 elements through the expected plastic hinge length (0.5d to d, where d = section depth).
- Verify material data range. The stress-strain curve must extend to the strain level expected. Truncated data causes the solver to extrapolate, often producing negative stiffness.
- Add artificial damping/stabilization. ABAQUS: *STATIC, STABILIZE=2e-5. SAP2000: increase Newton iterations per step to 100. Verify energy ratio stays below 5%.
- Validate against a simplified model. Build a single-element or single-frame version and confirm convergence. Then add complexity layer by layer.
9. Interactive: Convergence Parameter Calculator
Have Feedback?
Feel free to drop your comments below. I usually reply within 8 to 24 hours.
