AMD Numberer

The Approximate Minimum Degree (AMD) numberer implements the AMD algorithm to minimize fill-in during matrix factorization by optimizing the equation numbering scheme.

Description

The AMD numberer: - Uses graph theory to minimize matrix fill-in during factorization - Reorders DOFs to reduce computational cost - Particularly effective for sparse matrices - Implemented as a singleton to ensure consistent numbering across the model

Parameters

Parameter	Type	Description
None		This numberer doesn’t take any parameters

Usage Example

# Create a Femora instance


# Get an AMD numberer
fm.analysis.numberer.get_numberer("AMD")

Notes

• Optimizes matrix structure to minimize fill-in during factorization

• Particularly effective for irregular mesh patterns

• Can significantly reduce memory requirements and computational cost

• May require more setup time than Plain numberer

• Recommended for large-scale problems with irregular patterns

• Especially beneficial when using sparse direct solvers