Parallel efficient global optimization method: A novel approach for time-dependent reliability analysis and applications. (1st December 2021)

• Record Type: Journal Article
• Authors: Wu, Jiawei; Jiang, Zhenliang; Song, Huaming; Wan, Liangqi; Huang, Fu
• Is Part Of: Expert systems with applications. Volume 184(2021)
• Publication Date: 2021-12-01
• Subjects: EGO Efficient Global Optimization -- EI Expected Improvement -- MCS Monte Carlo Simulation -- M-EGO-TRA Mixed Efficient Global Optimization for Time-dependent Reliability Analysis -- NOF Number of Function -- O-EGO-TRA Ordinary Efficient Global Optimization for Time-dependent Reliability Analysis -- PEGO Parallel Efficient Global Optimization -- P-EGO-TRA Parallel Efficient Global Optimization for Time-dependent Reliability Analysis -- PEI Parallel Expected Improvement -- TRA Time-dependent Reliability Analysis; Time-dependent reliability analysis -- Failure probability -- Parallel efficient global optimization -- Kriging model -- Monte Carlo simulation; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2021.115494
• Languages: English
• ISSNs: 0957-4174

Abstract

Graphical abstract: Highlights: Parallel efficient global optimization and adaptive Kriging-Monte Carlo simulation were integrated. The proposed method outperformed the existing methods in computing efficiency. The proposed method showed an excellent accuracy in high-dimension problems. Abstract: Time-dependent failure may result from material properties deterioration, random load, and other uncertain factors, which are widespread among practical engineering applications. However, in previous studies of time-dependent reliability analysis (TRA), parallel computing, which can speed up the optimization process significantly, has not been fully considered. Therefore, this study proposed a parallel efficient global optimization (PEGO) integrated with the adaptive Kriging-Monte Carlo simulation (AK-MCS) for the TRA problems. It was shown that the proposed method was superior to the existing TRA methods in computing efficiency and maintained a high accuracy to solve the high-dimension of the TRA problems. Besides, it exhibited an extensive application scope due to the excellent performances (i.e., efficiency, accuracy), which were verified via six representative cases, including four low-dimension and two high-dimension, by comparing with the existing TRA methods.