MxL-GEN
MxL-GEN: A Modular Workflow for Mechanistic Learning based Surrogates, for use in Optimisation
MxL-GEN is a modular codebase implementing mechanistic learning: the combination of machine learning and mechanistic models.
The main sections o fthis codebase are as follows:
- High-fidelity ground truth simulations
- Fast reduced-order models
- Machine-learning surrogates
- Coupling to pyMOS optimiser (a multi-strategy evolutionary optimiser)
The framework is intended for problems where high-fidelity simulations are expensive or slow and a surrogate can greatly accelerate optimisation.
Download Link
Relevant publications
- E. Hayman, J.-M. Peña, J. Jayamohan, S. Waters and A. Jerusalem. Mechanistic learning based surrogate-aided generative design: Application to hydrocephalus shunt systems. Journal of Mechanical Design, in press