Settings for large projects
Settings for Large Projects in Twin Fabrica
As project complexity increases—especially with large, highly meshed geometries—the computational demands on the simulation engine also grow. In Twin Fabrica, special considerations and settings are recommended to ensure efficient and stable performance during Full-Order Model (FOM) simulation and Model Order Reduction (MOR).
This guide outlines best practices and configuration adjustments for working with large-scale models.
1. Assess Mesh Size
The total number of mesh elements directly affects:
- FOM solver performance
- Model parametrization speed
- MOR convergence and runtime
You can view mesh statistics in the PHYSICS tab, under the Physics Information panel on the right-hand side.
2. Access Solver Settings
For large models, solver behavior should be adjusted for better performance:
- Open the PHYSICS tab
- Click on Settings in the top menu bar
- Navigate to the Solver Configuration section
These settings allow you to choose between solver types and fine-tune stopping criteria.
3. Switch to Iterative Solver (Recommended)
By default, Twin Fabrica uses a direct solver, which is robust but memory-intensive.
For large models (especially those exceeding 500,000 elements), we recommend switching to the iterative solver, which is more scalable and memory-efficient.
This option is available directly within the solver settings window.
4. Configure Iterative Solver Parameters
The iterative solver uses two key parameters to control convergence:
- Tolerance – Defines the numerical precision target
- Maximum Iterations – Limits the time spent solving each system
You can adjust both values to balance accuracy and performance depending on your project needs.
5. Impact on FOM Simulation Time
Solver settings have a direct impact on FOM computation time:
- Relaxed tolerance values (e.g., 1e-3 or 1e-4) and reasonable iteration limits can reduce simulation time significantly
- Tighter tolerances (e.g., 1e-6 or lower) increase accuracy but may result in slower performance
6. Impact on MOR Runtime
Solver configuration also affects the Model Order Reduction phase.
During MOR, the full-order model is solved multiple times. Thus:
- Stricter tolerances lead to slower MOR due to more iterations
- Looser tolerances can accelerate reduction while still delivering acceptable ROM accuracy
7. General Recommendation
✅ For models with more than 500,000 mesh elements, use the iterative solver✅ Adjust tolerance and iteration limits based on the required accuracy
✅ Monitor solver behavior via logs to fine-tune for your hardware and project scale
Adapting these settings helps maintain responsiveness and reduces computational costs in large simulation projects.