Neural network-based chaotic crossover method for structural reliability analysis considering time-dependent parameters

Abstract

Structural reliability analysis with various time-dependent parameters is important for operational safety. To improve the calculation precision and efficiency for structural time-dependent reliability analysis, the neural network-based chaotic crossover method (CCM-NN) is proposed by absorbing extremum thought, artificial neural network, chaotic crossover strategy, reptile search algorithm (RSA), and Bayesian regularization (BR) algorithm. The availability of the CCM-NN method is validated by the time-dependent reliability analysis of aeroengine turbine blisk. The results show that (i) the developed CCM-NN method has superior modeling characteristics, whose modeling time and the average absolute error are 0.36 s and 2.26 × 10−4 m respectively; (ii) the CCM-NN methods holds eminent simulation feature, 0.247 s and 99.98% are simulation time and precision respectively since the Monte Carlo samples is 5 × 103; (iii) the time-dependent reliability of turbine blisk is 0.9987 when the allowable radial deformation of turbine blisk is 1.9215 × 10−3 m. The efforts of this study offer useful insight for structural reliability analysis by considering the effect of dynamic loads, and enrich mechanical reliability theory and method.