Premium
A Multi‐fidelity Approach Using Physical and Mathematical Surrogates for Crash Optimization
Author(s) -
Komeilizadeh Koushyar,
Hefele Rafael,
Duddeck Fabian
Publication year - 2018
Publication title -
pamm
Language(s) - English
Resource type - Journals
ISSN - 1617-7061
DOI - 10.1002/pamm.201800287
Subject(s) - fidelity , crash , computer science , high fidelity , context (archaeology) , mathematical optimization , optimization problem , nonlinear system , algorithm , mathematics , engineering , telecommunications , programming language , paleontology , physics , quantum mechanics , electrical engineering , biology
In the context of crash optimization, numerical simulations can be very time consuming, therefore efficient optimization algorithms are indispensable. If the original crash model is considered as the high‐fidelity (most accurate) one, low‐fidelity models trade accuracy for efficiency. A proper combination of high‐ and low‐fidelity models can lead to a more efficient optimization algorithm in total. We developed a new multi‐fidelity, or multi‐model optimization algorithm that is based on nonlinear crash simulations while being accelerated by gradients of the (initial) linear system used as low‐fidelity model.