Surrogate-modeling-assisted creep-fatigue reliability assessment in a low-pressure turbine disc considering multi-source uncertainty

This paper proposes a surrogate modeling approach based on XGboost machine learning technique, in order to establish a data-driven mapping relationship between input and output abstracted from practical finite element analysis (FEA) results. It facilitates novel insights into an efficient application of creep-fatigue reliability assessment in low-pressure turbine disk without a large amount of high-fidelity FEA cases. In detail, a general technical route is proposed for the probabilistic estimations of creep-fatigue lifetimes, where the multi-source uncertainties in the sequenced levels are synchronously considered. Subjected to typical creep-fatigue load spectrum, precise weakness hotspot is identified at the 1st bottom fir-tree groove of the turbine disk. Based on hotspot-based strategy, it is found that XGboost-involved surrogate modeling approach significantly improves the computational efficiency. The common results show that logarithmic creep-fatigue lifetimes roughly obey the normal distributions with the present of uncertainty sources, regardless of the multi-source combinations. Specifically, geometric tolerance plays an important role in reliability assessment results, which not only makes conservative gap but also shows high sensitivity in the reliability assessments.