Welcome to skrom

scikit-rom

scikit-rom is a lightweight, Python-based platform designed for projection-based model reduction of finite element models with moderate to large problem sizes (up to approximately 600,000 degrees of freedom). Built on top of scikit-fem, an easy-to-use finite element assembly library, this repository demonstrates reduced-order modeling (ROM) techniques for a range of linear and nonlinear problems, including time-dependent ones. Current examples primarily focus on thermal and mechanical systems, with extensions to fluid dynamics planned for future releases.

Beyond its application to real-world finite element problems, the library also serves as a valuable testing ground for new model reduction algorithms and as a tutorial resource for those new to the field.

What is Reduced-Order Modeling?

Reduced-order modeling (ROM) is a computational strategy for approximating high-fidelity numerical models with significantly reduced computational cost. Instead of solving large-scale systems—often with millions of unknowns—ROM techniques construct low-dimensional surrogates that retain the dominant physics of the system, enabling efficient and repeated evaluations.

The underlying principle is that many physical systems exhibit behavior that can be captured by a small number of dominant modes, despite their discretization on fine spatial meshes. Techniques such as Proper Orthogonal Decomposition (POD) extract these modes from full-order simulations, enabling the construction of reduced models that are both accurate and computationally efficient.

Offline–Online Decomposition

scikit-rom implements the classical offline–online decomposition paradigm:

  • Offline Phase (computationally intensive, performed once):

    • Sampling of the parameter space
    • Generation of full-order solution snapshots
    • Computation of the reduced basis via singular value decomposition (SVD)
    • Assembly of parameter-independent reduced operators

  • Online Phase (efficient, performed repeatedly):

    • Evaluation of the system response for new parameter values using preassembled reduced operators
    • Speedups on the order of 10–500× relative to full-order models are typical

Installation

This tutorial explains how to install SKROM (scikit-ROM):

Installation Steps:

  1. Clone the repository:
git clone https://github.com/suparnob100/scikit-rom
cd scikit-rom
  1. Install with pip:
  • For developers (editable install):
pip install -e .
  • For regular users:
pip install .

(Use pip3 instead of pip on Mac)

Notes:

  • Use -e for development; it lets you edit the code without reinstalling.
  • Regular users should use standard installation.
  • Reinstallation automatically removes the old version.
  • The installation completes in seconds and shows “successfully built SKROM.”
  • All instructions are available on the GitHub page.

Getting Started