ITEA is the Eureka Cluster on software innovation
ITEA is the Eureka Cluster on software innovation
Please note that the ITEA Office will be closed from 25 December 2024 to 1 January 2025 inclusive.
ITEA 4 page header azure circular

Implementation of a Krylov reduction algorithm into Simcenter ROM Builder

Project
19037 COMPAS
Type
Enhancement
Description

A novel multifrequency implementation of the Krylov method has been explored and implemented into the commercial software "Simcenter ROM builder" to automate the ROM creation process for mechanical and thermal systems. Simcenter ROM Builder is a general-purpose model order reduction tool. It enables creation and export of reduced order models from various sources. The new method demonstrates substantial improvement over simpler single frequency Krylov algorithm (see delivrable 4.1)., and its integration within ROM-builder give users an easy to use tool to both generate ROM and export them.

Contact
Gwendal Jouan, Siemens Industry Software
Email
gwendal.jouan@siemens.com
Research area(s)
reduced order models
Technical features

A novel multifrequency implementation of the Krylov method has been explored to automate the ROM creation process for both mechanical and thermal systems. The Krylov method is based on sampling solution vectors and their derivatives at different frequencies and creating a reduction basis that is tailored for a given input/output dof set. The difficulty is in automating this process, such as how many vectors to create per frequency, how many frequencies to sample, and so on. Previous work [1] has explored this for an application to second order systems (mechanical). It is based on comparing the responses of two consecutive ROMs and checking convergence based on the target tolerance value. Firstly, the method has been extended to first order systems (thermal or coupled transient thermal and static mechanical), for which an iterative solver implementation has been done. While for second order dynamic systems, such solvers may experience convergence issues, for first order systems, they work robustly and efficiently. An additional speed up is achieved by generating the initial guess of the iterative solver for each new calculation. As the ROM is created adaptively, by step-by-step addition of basis vectors, it is possible to use the previous ROM to make a fast initial guess for the iterative solver. Overall, these improvements made tackling of very large thermal systems possible with the Krylov method.

Integration constraints

Simcenter ROM Builder is available both for Windows and Linux

Targeted customer(s)

Existing customers of Simcenter ROM Builder

Conditions for reuse

Licensed customers

Confidentiality
Public
Publication date
21-11-2023
Involved partners
NXP Semiconductors (NLD)
Fraunhofer ENAS (DEU)
Delft University of Technology (NLD)
Eindhoven University of Technology (NLD)
Siemens Industry Software NV (BEL)
MicroConsult Engineering GmbH (DEU)
Infineon Technologies AG (DEU)
Jade University of Applied Sciences (DEU)
Katholieke Universiteit Leuven (BEL)