PME-toolkit¶

Overview¶

PME-toolkit is an open-source framework for design-space dimensionality reduction based on the Parametric Model Embedding (PME) methodology and its physics-aware extensions.

The toolkit supports reproducible workflows for engineering design optimization and reduced-order modeling.

Key Features¶

reduced-order representation of design spaces
analytical backmapping to original variables
support for physics-informed (PI-PME) and physics-driven (PD-PME) variants
unified MATLAB and Python implementations
standardized input/output configuration (JSON-based workflows)

Mode 1

Mode 2

Mode 3

Example of first three PI-PME modes (glider dataset)

Capabilities¶

PME-toolkit enables:

dimensionality reduction of high-dimensional design spaces
integration of geometry, design variables, and physics
explainable surrogate-based optimization
reproducible benchmarking workflows

Resources¶

🔗 GitHub Repository
📖 Documentation
📦 PyPI
📄 Zenodo
📝 JOSS submission: under review

PME-toolkit is designed to work with benchmark datasets available in this repository:

Bio-inspired underwater glider
RAE2822 airfoil

👉 See Datasets

Citation¶

If you use PME-toolkit in your research, please cite the associated publication (JOSS or related articles).

Development Status¶

actively developed
tested on benchmark datasets
aligned MATLAB/Python implementations