Efficient Surrogate Model-Assisted Evolutionary Algorithm for Electromagnetic Design Automation with Applications
AbstractIn this thesis, the surrogate model-aware evolutionary search (SMAS) framework is extended for efficient interactive optimisation of multiple criteria electromagnetic (EM) designs and/or devices through a novel method called two-stage interactive efficient EM micro-actuator design optimisation (TIEMO). The first robust analytical and behavioural study of the SMAS framework is also carried out in this thesis to serve as a guide for the meticulous selection of multiple differential evolution (DE) mutation strategies to make SMAS fit for use in parallel computing environments. Based on the study of SMAS and the self-adaptive use of the selected multiple DE mutation strategies and reinforcement learning techniques, a novel method, parallel surrogate model-assisted evolutionary algorithm for EM design (PSAED) is proposed. PSAED is tested extensively using mathematical benchmark problems and numerical EM design problems. For all cases, the efficiency improvement of PSAED compared to state-of-the-art evolutionary algorithms (EAs) is demonstrated by the several times up to about 20 times speed improvement observed and the high quality of design solutions. PSAED is then applied to real-world EM design problems as two purposebuilt methods for antenna design and optimisation and high-performance microelectro-mechanical systems (MEMS) design and optimisation in parallel computing environments, parallel surrogate model-assisted hybrid DE for antenna optimisation (PSADEA) and adaptive surrogate model-assisted differential evolution for MEMS optimisation (ASDEMO), respectively. For all the real-world antenna and MEMS design cases, PSAED methods obtain very satisfactory design solutions using an affordable optimisation time and comparisons are made with available alternative methods. Results from the comparisons show that PSAED methods obtain very satisfactory design solutions in all runs using an affordable optimisation time in each, whereas the alternative methods fail and/or seldom succeed to obtain feasible or satisfactory design solutions. PSAED methods also show better robustness and stability. In the future, PSAED methods will be embedded into commercial CAD/CEM tools and will be further extended for use in higher-order parallel clusters.
CitationAkinsolu, M, O. (2019). Efficient Surrogate Model-Assisted Evolutionary Algorithm for Electromagnetic Design Automation with Applications (Doctoral dissertation). University of Chester, UK.
PublisherUniversity of Chester
TypeThesis or dissertation
The following license files are associated with this item:
- Creative Commons
Except where otherwise noted, this item's license is described as Attribution-NonCommercial-NoDerivatives 4.0 International