Active extremum Kriging-based multi-level linkage reliability analysis and its application in aeroengine mechanism systems

To improve the computational efficiency and accuracy of dynamic multi-component system reliability analysis involving complex characteristics like dynamic traits, high-nonlinearity, and failure correlation, an active extremum Kriging-based multi-level linkage method (AEK-MLL) is proposed by incorporating the benefits of extremum selection technique, active learning Kriging model, and the multi-level linkage strategy. In AEK-MLL modeling, the extremum selection technique first converts the dynamic candidate sample domain into a steady-state candidate sample domain, and the active learning technique searches for the best training samples, to build the active extremum Kriging model of component-level limit state functions (LSFs); moreover, the multi-level linkage strategy is adopted to take failure correlation into account, to establish a reliability framework for complex dynamic multi-component systems. The proposed method is first validated by a dynamic numerical case and three system numerical cases, and then applied to the dynamic multi-component reliability analysis of a typical aero-engine mechanism system. The numerical cases and engineering case show that the AEK-MLL holds the high-accuracy and high-efficiency in dealing with the complex dynamic multi-component system reliability analysis.(c) 2022 Elsevier Masson SAS. All rights reserved.