Structural dynamic probabilistic evaluation using a surrogate model and genetic algorithm

Abstract

To improve the computational accuracy and efficiency of dynamic probabilistic analysis of complex structures, an extremum Kriging method (EKM) with multi-population genetic algorithm (MPGA) is proposed fusing extremum response surface method (ERSM), Kriging model and MPGA. Particularly, the EKM is developed by combining the Kriging model and ERSM to handle dynamic processes of related variables and to reduce computational burden by regarding the extreme value of response process in each dynamic analysis within the time domain. The MPGA is used to replace gradient decent to find the hyperparameter θ in the Kriging model by solving the maximum-likelihood equation. The effectiveness of the proposed method was validated by performing the dynamic probabilistic analysis of an aeroengine high-pressure compressor blisk radial running deformation with fluid-thermal-structural interaction. The analytical results illustrate that the reliability degree of the blisk is 0·9956 under the allowable value uallow = 1·75 × 10−3 m, and gas temperature is the leading factor against output response, followed by rotational speed, inlet velocity, material density and outlet pressure. Moreover, the developed MPGA-EKM is superior to other methods in computational accuracy and efficiency. The efforts of this study provide a useful insight to design complex structures and enrich mechanical reliability theory.