/al_cfd_benchmark

Active learning based regression for CFD cases

Primary LanguageC++MIT LicenseMIT

active_learning_cfd

Active learning based regression for CFD cases

Requirements

This requires the following packages and their dependencies:

  • numpy
  • matplotlib
  • modAL
  • PyFoam

To run the examples, it is also needed:

  • sklearn

Installation

We recommend that the package be installed in development mode:

pip3 install -e .

Usage

Examples cases are provided in the example folder.

The test cases presented on the article are available on the cases folder:

  1. static_mixer
  2. orifice
  3. mixer
  4. mixer3D

Each case is composed by a folder with the OpenFOAM template and a runner python script for running the case and extracting outputs. The regression_batch scripts run a set of different strategies, with the possibility of repeating each one several times for statistics. The reference scripts generate reference results for estimation of interpolation error.

About

G. F. N. Gonçalves, A. Batchvarov, Y. Liu, Y. Liu, L. Mason, I. Pan, O. K. Matar (2020). Data-driven surrogate modelling and benchmarking for process equipment. Data-Centric Engineering. DOI: 10.1017/dce.2020.8