Ansys Nonlinear Convergence: Fixing 'Solver Engine Unable to Converge' Errors

Ansys Nonlinear Convergence: Fixing "Solver Engine Unable to Converge" Errors

A user on the Ansys Learning Forum described a problem I've encountered countless times: they were designing a brake rotor and running analysis with frictional contacts. After a few minutes, the solver would stop with warnings about friction values greater than 0.2 and an error stating the solver engine was unable to converge on a solution. Another user reported convergence errors while simulating a 2D circle with large displacement at the top edge and fixed bottom edge. A third user struggled with a frictional contact between a steel support structure and a tungsten wire — the analysis simply refused to converge.

Nonlinear convergence failure is the most common problem in Ansys Mechanical, and it's also the most frustrating because there's no single fix. The solver's inability to converge means the Newton-Raphson iterations aren't finding equilibrium — the internal forces don't balance the external forces within the tolerance. I'll walk through the systematic approach I use to diagnose and fix convergence problems.

Understanding Nonlinear Convergence

In a linear static analysis, the solver solves a single equation: [K]{u} = {F}. In a nonlinear analysis, the stiffness matrix [K] changes with displacement, so the solver uses Newton-Raphson iterations:

1. Apply a load increment
2. Estimate the displacement
3. Calculate internal forces based on current geometry and material state
4. Compare internal forces to external forces (residual)
5. If residual > tolerance, adjust displacement and repeat
6. If residual < tolerance, the substep has converged — move to the next load increment

Convergence failure means step 5 keeps failing — the residual never drops below the tolerance. This happens for several reasons, and identifying which one is the key to fixing it.

Step 1: Check the Newton-Raphson Residual Plot

Ansys can plot the residual force distribution, which shows where the imbalance is concentrated.

1. In Mechanical, go to Solution Information
2. Set Solution Output to Solver Output
3. Look for the NR residual convergence graph
4. If the residual is oscillating or increasing, the solver is diverging
5. Right-click Solution Information → Insert → Newton-Raphson Residual
6. Set the number of residual plots to 3-5
7. Re-run the analysis
8. The residual plots show where the force imbalance is concentrated

The location of the residual tells you what's causing the problem:

• At a contact interface: Contact stiffness or friction issue
• At a material transition: Material model issue
• At a constraint: Over-constraint or rigid body motion
• At a sharp corner: Stress singularity

Step 2: Increase the Number of Substeps

The most common cause of convergence failure is trying to apply too much load in a single step. The Newton-Raphson algorithm works best with small load increments.

1. In the analysis settings, go to Step Controls
2. Set Number of Substeps to a larger value:
   • Start with 100 initial substeps
   • Set minimum substeps to 10
   • Set maximum substeps to 1000
3. Set Auto Time Stepping to On
4. The solver will automatically adjust the substep size based on convergence behavior

For Large Displacement Problems

If you have large displacements (common in metal forming, rubber analysis, or buckling):

1. Analysis Settings → Large Deflection: Set to On
2. Increase substeps to 200-500 initial
3. The solver needs smaller increments because the geometry changes significantly during loading

A forum user reported convergence errors with large deflection enabled on a 2D circle problem. The fix was to increase the number of substeps and enable auto time stepping, allowing the solver to take smaller increments when needed.

Step 3: Adjust Contact Stiffness

Frictional contacts are the most common source of convergence problems. The contact stiffness determines how "hard" the contact is — too stiff and the solver bounces, too soft and the surfaces penetrate.

Normal Stiffness

1. Go to the Connections branch in the tree
2. Select the frictional contact
3. In the Properties:
   • Normal Stiffness: Change from Program Controlled to Manual
   • Set Normal Stiffness Factor (FKN): Start at 0.1 (softer than default)
   • If penetration occurs, increase to 1.0
   • If convergence fails due to bouncing, decrease to 0.01

Update Stiffness

1. Update Stiffness: Set to Each Iteration (default is Each Substep)
2. This allows the contact stiffness to adapt more frequently
3. It's slower but more robust for difficult contact problems

For Friction Values > 0.2

The forum user's brake rotor analysis triggered a warning about friction values greater than 0.2. This warning appears because high friction coefficients make the contact more nonlinear — the tangential forces are larger, and the Newton-Raphson iterations have more difficulty finding equilibrium.

Fixes:

1. Reduce the friction coefficient slightly (e.g., from 0.4 to 0.35) to see if convergence improves
2. If the physical friction is genuinely high, use the MPC (Multipoint Constraint) contact algorithm instead of the penalty-based algorithm
3. Or use the Pure Penalty algorithm with a lower stiffness factor

Step 4: Check Material Models

Nonlinear material models (plasticity, hyperelasticity, creep) can cause convergence problems if they're not properly configured.

Plasticity (Bilinear or Multilinear Isotropic Hardening)

1. Check the stress-strain curve — ensure it's monotonically increasing
2. For multilinear hardening, ensure the points are in ascending order of strain
3. If the material yields and then softens, the solver may not converge (unless you use the Hill model or damage mechanics)

Hyperelasticity (Mooney-Rivlin, Ogden, Yeoh)

1. Ensure the material constants are stable — bad curve fitting can produce unstable models
2. Use the Mixed u-P formulation: Analysis Settings → Solver Formulation → Mixed u-P
3. This formulation is more robust for nearly incompressible materials (rubber, biological tissue)

Check for Material Instability

1. Run a simple uniaxial tension test in Ansys with your material model
2. If the single-element test doesn't converge, your material model is the problem
3. Fix the material constants before running the full analysis

Step 5: Improve Mesh Quality

Poor mesh quality causes convergence problems because distorted elements have poor conditioning, making the stiffness matrix harder to solve.

Check Mesh Metrics

1. Go to Mesh → Quality
2. Check:
   • Element Quality: Should be > 0.1 (1.0 is perfect)
   • Aspect Ratio: Should be < 20 for most elements
   • Jacobian Ratio: Should be > 0.5 (1.0 is perfect)
   • Skewness: Should be < 0.85

Refine Mesh at Contact Interfaces

The contact region needs a fine mesh to accurately capture contact pressure distribution:

1. Add a Body Sizing control on the contact surfaces
2. Set the element size to 1/10 of the contact width
3. Use Hex Dominant meshing for the contact region if possible
4. Ensure the mesh on both sides of the contact is similar in size (avoid large size ratios)

Step 6: Use the Right Solver

Ansys offers two main solvers:

• Direct (Sparse) Solver: More robust, handles ill-conditioned matrices better, but uses more RAM
• Iterative (PCG) Solver: Faster for large models with good mesh quality, but less robust

For convergence problems:

1. Analysis Settings → Solver Type: Switch to Direct
2. The direct solver is more likely to converge when the stiffness matrix is poorly conditioned
3. If the direct solver runs out of memory, use the Iterative solver with a good mesh

Step 7: Stabilize with Damping

If the structure is unstable (rigid body motion, snap-through buckling, contact separation), adding artificial damping can help the solver converge:

1. Analysis Settings → Stabilization: Set to Constant
2. Set Stabilization Energy Factor: 1e-4 (start small)
3. This adds a small damping force that prevents instability
4. Check the stabilization energy in the results — it should be less than 5% of the strain energy
5. If stabilization energy is too high, reduce the factor and re-run

Step 8: Simplify the Model

If all else fails, simplify the model to identify the problematic feature:

1. Remove frictional contacts and replace with bonded contacts — does it converge?
2. Remove nonlinear materials and use linear elastic — does it converge?
3. Remove large deflection — does it converge?
4. By removing one nonlinearity at a time, you can identify which feature is causing the convergence failure

Summary

Nonlinear convergence is a process of elimination. Start with the most common fixes and work toward the less common ones:

| Fix | Success Rate | Difficulty | |-----|-------------|------------| | Increase substeps | 40% | Easy | | Adjust contact stiffness | 25% | Easy | | Enable large deflection | 10% | Easy | | Improve mesh quality | 10% | Medium | | Fix material model | 5% | Medium | | Switch to direct solver | 5% | Easy | | Add stabilization | 3% | Easy | | Simplify model | 2% | Medium |

The Newton-Raphson residual plot is your most important diagnostic tool — it shows you exactly where the convergence problem is located. Start there, and target your fix to the identified region.