A relative adequacy framework for multimodel management in multidisciplinary design optimization
- PDF / 5,238,998 Bytes
- 20 Pages / 595.224 x 790.955 pts Page_size
- 9 Downloads / 142 Views
RESEARCH PAPER
A relative adequacy framework for multimodel management in multidisciplinary design optimization Ahmed H. Bayoumy1 · Michael Kokkolaras1 Received: 15 October 2019 / Revised: 4 March 2020 / Accepted: 28 March 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract In previous work, we presented a novel relative adequacy framework to manage the employment of a set of available computational models in (single-disciplinary) design optimization problems. In this paper, we extend our method to solve multidisciplinary design optimization problems with particular emphasis on strongly coupled fluid-structure interactions. We illustrate that these interactions can have a significant impact on multimodel management: models that may be selected in a single-disciplinary analysis context can be inadequate in a multidisciplinary analysis one. We implement our method for two multidisciplinary design optimization architectures: the monolithic multidisciplinary feasible formulation and a penaltybased distributed interdisciplinary feasible formulation. We illustrate the proposed multimodel management methodology by means of two example problems: a flexible beam fluid-structure interaction problem and a transonic fan flow problem. The obtained results demonstrate that our framework is accurate and efficient while exhibiting significant computational cost benefits, especially when disciplinary coupling is tight. Keywords Relative adequacy framework · Multidisciplinary design optimization · Multidisciplinary analysis · Fluid-structure interaction
1 Introduction The field of multifidelity optimization (MFO) is evolving toward approaches capable of leveraging the synergistic employment of physics- and data-based models. In multidisciplinary design optimization (MDO), the type and strength of disciplinary coupling can have a significant impact on multimodel management. In previous work, we introduced a relative adequacy framework (RAF) for managing the use of models (whether physics- or data-based) of varying fidelity regardless of their disciplinary context (Bayoumy and Kokkolaras 2019). In this paper, we extend our framework to solve MDO problems that involve tight coupling (analysis feedback loops). Interdisciplinary couplings can Responsible Editor: Nathalie Bartoli Michael Kokkolaras
[email protected] Ahmed H. Bayoumy [email protected] 1
Department of Mechanical Engineering, McGill University, Montreal, Quebec, H3A 0C3, Canada
be categorized as being “tight” or “loose” based on the response of a discipline output to a change in a coupling variable (Allison et al. 2005; Baptista et al. 2018). Engineering systems have become more complex and, therefore, more challenging to design and manage (Simpson and Martins 2011). Engineering design of systems such as aircraft, automobiles, or gas turbines requires the solution of large-scale and multiphysics MDO problems involving multiple interacting disciplines. Simpson and Martins emphasized the need of multifidelity modeling for MD
Data Loading...
