TOPSIS based multi-fidelity Co-Kriging for multiple response prediction of structures with uncertainties through real-time hybrid simulation. (1st April 2023)

• Record Type: Journal Article
• Authors: Chen, Cheng; Ran, Desheng; Yang, Yanlin; Hou, Hetao; Peng, Changle
• Is Part Of: Engineering structures. Volume 280(2023)
• Publication Date: 2023-04-01
• Subjects: Real-time hybrid simulation -- Kriging -- Co-Kriging -- Sequential -- TOPSIS -- Uncertainty quantification -- Experimental design; Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2023.115734
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: Innovatively integrate real-time hybrid simulation with multi-fidelity co-Kriging meta -modeling for multiple response prediction with the presence of structural uncertainties. Integrate the entropy-based sequential sampling with TOPSIS to sequentially determine new sampling points for HF and LF simulation. RTHS of a two-degree-of-freedom system with self-centering viscous dampers are conducted to experimentally demonstrate the effectiveness of proposed approach. The effectiveness of proposed approach is further demonstrated through comparison with Kriging meta -modeling and Co-kriging meta -modeling without TOPSIS. Abstract: Energy dissipation devices in vibration control often present challenges for accurate modeling and uncertainty quantification through computational simulation. Simplified numerical models of these devices might not realistically represent their behavior under earthquakes thus lead to errors in response prediction and uncertainty quantification. This study further explores the integration of Co-Kriging meta -modeling and real-time hybrid simulation (RTHS) for global response prediction of multi-degree-of-freedom systems under the presence of structural uncertainties. RTHS in laboratory is taken as high-fidelity (HF) model while computational simulation with approximate modeling is used as low-fidelity (LF) model. Multi-fidelity modeling is integrated through Co-Kriging to render accurate response prediction over the entire sample space of uncertainty. An entropy-based sequential sampling is integrated with the T echnique for O rder of P reference by S imilarity to I deal S olution (TOPSIS) to sequentially determine new sampling points for HF and LF simulation. The proposed TOPSIS based multi-fidelity Co-Kriging approach is experimentally evaluated through RTHS of a two-degree-of-freedom structure with self-centering viscous dampers. Accuracy of Co-Kriging prediction are further evaluated through validation tests. It is demonstrated that TOPSIS can effectively reduce the number of RTHS tests in laboratory required by multi-fidelity Co-Kriging to achieve better prediction accuracy. The study presents an innovative and effective way to apply RTHS for efficient uncertainty quantification of multiple response quantities.