Fatigue reliability estimation framework for turbine rotor using multi-agent collaborative modeling

To improve the computing accuracy and simulation efficiency of fatigue reliability estimation for turbine rotor, a multi-agent collaborative modelling (MACM) approach is proposed by absorbing the strengths of improved differential evolution (IDE) algorithm and neural network model into decomposed-collaborative strategy. The fatigue reliability estimation framework is presented in respect of MACM approach. Furthermore, the fatigue reliability estimation of a typical turbine rotor is considered as engineering case to evaluate the presented approach with respect to material variabilities, load fluctuations and model randomness. The estimation results reveal that the probabilistic fatigue life of a turbine rotor under reliability 99.87% is 6 050 cycles; fatigue strength coefficient sigma(f)' and strain range Delta epsilon(t) play a leading role on the fatigue life since their effect probabilities of 45% and 36%, respectively. The comparisons of methods (direct Monte Carlo simulation (MCS), quadratic polynomial (QP), neural network (NN), neural network agent (NNA)) show that the presented MACM approach holds high efficiency and accuracy for fatigue reliability estimation of turbine rotor.

Automated summary

The proposed MACM approach combines the strengths of improved differential evolution algorithm and neural network model to improve the computing accuracy and simulation efficiency of fatigue reliability estimation for turbine rotor. Through the engineering case study of a typical turbine rotor, the method shows high efficiency and accuracy in considering material variabilities, load fluctuations, and model randomness.