Dynamic surrogate modeling approach for probabilistic creep-fatigue life evaluation of turbine disks

Abstract

To improve the modeling accuracy and simulation efficiency of probabilistic creep-fatigue life evaluation, a decomposed collaborative time-variant Kriging surrogate model (DCTKS) is proposed by absorbing the strengths of extremum selection technique and Kriging model into decomposed collaborative strategy. The probabilistic creep-fatigue life evaluation of a typical turbine disk is considered as one case to evaluate the proposed DCTKS method with respect to fluctuation of transient loads, nonlinearity of material properties and variability of models. In respect of this study, we find that the probabilistic creep-fatigue life of the turbine disk under the reliability degree of 0.998 7 is 946 cycles, and the transient fluctuating loads (body temperature and rotor speed) are the main factors of influencing creep-fatigue life. Through the comparison of methods (DCTKS, Monte Carlo simulation method, Kriging surrogate method, decomposed collaborative response surface method), the proposed DCTKS is demonstrated to possess the computational advantages in efficiency and accuracy for probabilistic creep-fatigue life evaluation.