Hello everyone,
I am modelling a 4.00 × 3.00 m masonry infill panel (200 mm thick) inside an elastic frame. The frame–panel interfaces use line-to-line rigid links with line hinges: free tangential sliding, no friction, and unilateral normal contact ("Fixed if negative vy").
Reference case: with a linear elastic panel and a horizontal load of 100 kN, the model runs without warnings (max. horizontal displacement 1.042 mm).
Problem case (model E2-18_M100_I1000, attached):
- Isotropic | Nonlinear Elastic (Surfaces/Solids), Mohr–Coulomb, Basic diagram
- E = 4500 MPa, ν = 0.25
- fy,c = 3.000 MPa; fy,t = 0.001 MPa
- "Linear elastic with nonlinear criteria" unchecked
- FE mesh 100 mm (1458 2D elements, 9312 equations)
- Geometrically linear, Newton–Raphson, 100 load increments, max. 1000 iterations, all precision/tolerance/robustness factors = 1.00
The calculation takes about 12 minutes. Several increments need hundreds of iterations, and at least increments 95, 96, 97, 98 and 100 reach the 1000-iteration limit without converging; the solver nevertheless proceeds to the next increment. The run ends with:
Warning 10037 — The required accuracy of the iteration calculation was not reached | The results may be unusable.
Final state: max. displacement ≈ 8.8 mm; horizontal reactions sum to 100 kN. The maximum principal stress in the panel is σ1 = +0.092 MPa, about 90 times the tensile limit, located at the corners; σ1,min = −1.84 MPa, below fy,c.
Mesh check: the same model with a 50 mm mesh (otherwise identical) also ends with warning 10037, after ≈ 2 h 20 min, with repeated displacement spikes and a final max. displacement of ≈ 15.4 mm. The mesh therefore does not seem to cause the non-convergence, but the results are clearly mesh-dependent.
Questions:
- Which convergence criterion fails, and where can I inspect the residual, the tolerance and which increments actually converged?
- Is it intended that RFEM proceeds to the next increment after hitting the iteration limit without convergence, warning only at the end? Can the calculation be set to stop at the first non-converged increment?
- Does σ1 = 0.092 MPa against fy,t = 0.001 MPa confirm that the final state violates the material criterion, or does the Basic diagram apply the limit to a different stress measure?
- Is this material model suitable for negligible tensile strength with a defined compressive limit and recovery of compressive transfer on reclosure (no accumulated damage)? If not, which model would you recommend?
- Since convergence is lost from about 95 % of the load onwards, how can I distinguish a numerical or contact convergence problem from the panel approaching its load capacity?
I would prefer to understand the cause before increasing iterations or relaxing tolerances.
Add-ons active: Nonlinear Material Behavior, Masonry Design
Thank you,
Rui Camposinhos
E2-18_M100_I1000.rf6 (3.4 MB)