Design of IMEXRK time integration schemes via Delaunay-based derivative-free optimization with nonconvex constraints and grid-based acceleration

Ryan Alimo (), Daniele Cavaglieri (), Pooriya Beyhaghi () and Thomas R. Bewley ()
Additional contact information
Ryan Alimo: UC San Diego
Daniele Cavaglieri: UC San Diego
Pooriya Beyhaghi: UC San Diego
Thomas R. Bewley: UC San Diego

Journal of Global Optimization, 2021, vol. 79, issue 3, No 2, 567-591

Abstract: Abstract This paper develops a powerful new variant, dubbed $$\varDelta $$ Δ -DOGS( $$\varOmega _Z$$ Ω Z ), of our Delaunay-based Derivative-free Optimization via Global Surrogates family of algorithms, and uses it to identify a new, low-storage, high-accuracy, Implicit/Explicit Runge–Kutta (IMEXRK) time integration scheme for the stiff ODEs arising in high performance computing applications, like the simulation of turbulence. The $$\varDelta $$ Δ -DOGS( $$\varOmega _Z$$ Ω Z ) algorithm, which we prove to be globally convergent under the appropriate assumptions, combines (a) the essential ideas of our $$\varDelta $$ Δ -DOGS( $$\varOmega $$ Ω ) algorithm, which is designed to efficiently optimize a nonconvex objective function f(x) within a nonconvex feasible domain $$\varOmega $$ Ω defined by a number of constraint functions $$c_\kappa (x)$$ c κ ( x ) , with (b) our $$\varDelta $$ Δ -DOGS(Z) algorithm, which reduces the number of function evaluations on the boundary of the search domain via the restriction that all function evaluations lie on a Cartesian grid, which is successively refined as the iterations proceed. The optimization of the parameters of low-storage IMEXRK schemes involves a complicated set of nonconvex constraints, which leads to a challenging disconnected feasible domain, and a highly nonconvex objective function; our simulations indicate significantly faster convergence using $$\varDelta $$ Δ -DOGS( $$\varOmega _Z$$ Ω Z ) as compared with the original $$\varDelta $$ Δ -DOGS( $$\varOmega $$ Ω ) optimization algorithm on the problem of tuning the parameters of such schemes. A low-storage third-order IMEXRK scheme with remarkably good stability and accuracy properties is ultimately identified using this approach, and is briefly tested on Burgers’ equation.

Keywords: Derivative-free global optimization; Nonconvex constraints; IMEXRK time marching schemes; Computational fluid dynamics (search for similar items in EconPapers)
Date: 2021
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10898-019-00855-1 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:jglopt:v:79:y:2021:i:3:d:10.1007_s10898-019-00855-1

Ordering information: This journal article can be ordered from
http://www.springer. ... search/journal/10898

DOI: 10.1007/s10898-019-00855-1

Access Statistics for this article

Journal of Global Optimization is currently edited by Sergiy Butenko

More articles in Journal of Global Optimization from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().