A modified trust-region assisted variable-fidelity optimization framework for computationally expensive problems
ISSN: 0264-4401
Article publication date: 6 May 2022
Issue publication date: 5 July 2022
Abstract
Purpose
Variable-fidelity optimization (VFO) frameworks generally aim at taking full advantage of high-fidelity (HF) and low-fidelity (LF) models to solve computationally expensive problems. The purpose of this paper is to develop a novel modified trust-region assisted variable-fidelity optimization (MTR-VFO) framework that can improve the optimization efficiency for computationally expensive engineering design problems.
Design/methodology/approach
Though the LF model is rough and inaccurate, it probably contains the gradient information and trend of the computationally expensive HF model. In the proposed framework, the extreme locations of the LF kriging model are firstly utilized to enhance the HF kriging model, and then a modified trust-region (MTR) method is presented for efficient local search. The proposed MTR-VFO framework is verified through comparison with three typical methods on some benchmark problems, and it is also applied to optimize the configuration of underwater tandem wings.
Findings
The results indicate that the proposed MTR-VFO framework is more effective than some existing typical methods and it has the potential of solving computationally expensive problems more efficiently.
Originality/value
The extreme locations of LF models are utilized to improve the accuracy of HF models and a MTR method is first proposed for local search without utilizing HF gradient. Besides, a novel MTR-VFO framework is presented which is verified to be more effective than some existing typical methods and shows great potential of solving computationally expensive problems effectively.
Keywords
Acknowledgements
The authors would like to thank the colleagues and reviewers for their valuable comments.
Funding: This research was financially supported by the China Postdoctoral Science Foundation (Grant No. 2021M700529), the Natural Science Basic Research Program of Shaanxi Province (Grant No. 2022JQ-434), the Fundamental Research Funds for the Central Universities, CHD (Grant No. 300102252110) and the National Natural Science Foundation of China (Grant No. 51805436).
Authors' contributions: Huachao Dong studied the method and proposed modification strategies. Chengshan Li tested the method with benchmark problems and application problems and was also the major contributor in writing the manuscript. All authors read and approved the final manuscript.
Conflict of interest: On behalf of all authors, the corresponding author states that there is no conflict of interest.
Citation
Li, C. and Dong, H. (2022), "A modified trust-region assisted variable-fidelity optimization framework for computationally expensive problems", Engineering Computations, Vol. 39 No. 7, pp. 2733-2754. https://doi.org/10.1108/EC-08-2021-0456
Publisher
:Emerald Publishing Limited
Copyright © 2022, Emerald Publishing Limited