Cascade ensemble learning for multi-level reliability evaluation

doi:10.1016/j.ast.2024.109101

Cascade ensemble learning for multi-level reliability evaluation

2024-05-01

27

PRE

AI

DOI：10.1016/j.ast.2024.109101

Original

Original request for help

Save

Abstract

En 中文

For complex systems involving multiple operating conditions and multiple failure modes, its reliability analysis usually presents the cascade failure correlation between multiple levels (i.e., operating condition level, failure mode level) and the strong coupling analysis between multiple physical fields (i.e., fluid, thermal, structure), leading to the traditional integral or separate reliability modeling methods prone to unacceptable computing efficiency or accuracy. In this study, to improve the computing performance of the multi-level reliability analysis, by fusing the cascade synchronous strategy (CSS) and wavelet neural network-based AdaBoost (WNN-Ada) ensemble learning, a cascade ensemble learning (CEL) method is proposed. The complex composite function approximation (including three composite levels and nine nonlinear sub-functions) and the multi-level reliability evaluation of aeroengine turbine rotor system (including three operating conditions and three failure modes) are served as the numerical and engineering experiments, to evaluate the effectiveness of the present cascade ensemble learning. The comparison investigation in two experiments reveals that the presented approach shows evident advantages in terms of computing accuracy as well as computing efficiency. The current efforts shed light on the development of reliability modeling for complex multi-level systems.

Keywords:

System reliability

Reliability analysis

Multiple failure

Ensemble learning

Neural network

Cascade ensemble learning for multi-level reliability evaluation

Abstract

Journal

Organization