mosaictools 0.2.0

Surrogate modelling of modal properties using MOSAIC method

mosaictools

A Python package for surrogate modeling of modal properties using the Mode-Shape-Adapted Input parameter domain Cutting (MOSAIC) method.

Overview

The package is designed for structural dynamics workflows where mode degeneration effects (for example mode crossing, veering, or coalescence) make direct surrogate modeling difficult. MOSAIC addresses this by splitting the parameter domain into modal subdomains and fitting local generalized Polynomial Chaos Expansion (gPCE) models.

Reference preprint: MOSAIC method

Installation

Install from PyPI:

pip install mosaictools

Features

• Global multi-mode surrogate model through Mosaic.
• Automatic mode-wise subdomain discovery from eigenvector similarity (MAC-based clustering).
• Local frequency and eigenvector approximation in each subdomain.
• Built-in classifier support:
  • classification_method='svc' (scikit-learn SVC)
  • classification_method='custom' (any compatible classifier instance)
• K-fold cross-validation helper and modal error metrics.
• Model persistence to .msic files.

Demo Notebook and Data

A complete demonstration is included in:

• https://github.com/blazkurent/mosaictools/blob/main/notebooks/mosaic_package_demonstration.ipynb

The corresponding example arrays are in:

• https://github.com/blazkurent/mosaictools/tree/main/mosaic_example_data

License

This project is licensed under the MIT License. See LICENSE.