Demystifying Numerical Parameters in SimScale: A Guide for Advanced Users

 SimScale offers various numerical parameters for users who want to fine-tune their static analysis simulations. While these settings might seem complex, understanding their role can significantly enhance your control over the simulation process. Here's a breakdown of some key parameters:

Solver:

This defines the computational algorithm used to solve the system of equations generated by the finite element analysis (FEA) process. Common choices include:

• Direct Solver: Solves the equation system directly, ideal for smaller models.
• Iterative Solver: Solves the system iteratively, more efficient for larger models.

Precision:

• Single Precision: Uses 32-bit floating-point numbers, offering a balance between speed and accuracy for most simulations.
• Double Precision: Uses 64-bit floating-point numbers, providing higher accuracy for complex models or cases requiring very precise results.

Singularity Detection & Stop If Singular:

• Singularity Detection: Identifies situations where the stiffness matrix becomes singular (cannot be inverted), often indicating modeling errors or issues with boundary conditions.
• Stop If Singular: Defines whether the simulation should halt if a singularity is detected. Disabling this allows the simulation to continue, but results might be inaccurate.

Matrix Type:

• Sparse Matrix: Stores only non-zero elements of the stiffness matrix, saving memory for large models.
• Dense Matrix: Stores all elements, regardless of value, which might be faster for smaller models but consumes more memory.

Memory for Pivoting:

Allocates memory specifically for the pivoting process used by some solvers to improve stability. Increasing this value can help with convergence issues in certain cases.

Linear System Relative Residual:

Sets the convergence criterion for the iterative solver. The simulation stops when the relative difference between solution steps falls below this threshold. A lower value indicates higher accuracy but might require more iterations.

Preprocessing:

• Renumbering Method: This optimizes the order of equations in the system, potentially improving solver performance. Different methods offer varying trade-offs between speed and memory usage.

Postprocessing:

• Distributed Matrix Storage: This distributes the stiffness matrix across multiple compute cores for larger models, improving memory efficiency and potentially speeding up calculations.

Memory Management:

• Automatic Memory Management: SimScale automatically manages memory allocation for most users. However, advanced users can adjust settings for specific needs.

Important Note:

Modifying these parameters is recommended for experienced users who understand their impact on simulation accuracy, convergence, and resource usage. SimScale provides default settings optimized for most cases. It's best to consult the SimScale documentation or support for detailed information on each parameter before making adjustments.