Active learning Kriging approach for creep-fatigue reliability assessment of turbine disk

The active metamodel is hard to represent the creep-fatigue failure, which hinders the application of efficient active metamodeling technique in creep-fatigue reliability estimation. To improve the computing efficiency and accuracy of creep-fatigue reliability assessment for turbine disk involving complex coupling of multi-layer, multi-disciplinary, and multi-uncertainty, the efficient active metamodeling technique is first pushed deep into complex creep-fatigue reliability evaluation. By integrating the synergic surrogate strategy into the active metamodeling, a multi-layer surrogate control-based synergic enhanced Kriging (MSC-SEK) approach is proposed: Firstly, to precisely describe the complicated creep-fatigue strong-coupling relationships, a synergic enhanced Kriging (SEK) is established by organically synergizing multiple Kriging models, where a multi-colony multi-mutation artificial bee colony algorithm is designed to enhance the Kriging surrogate quality; further, to obtain high-quality modeling dataset, a novel MSC learning function is developed by synthetically considering multi-surrogate entropy and reliability-sensitive information. The superiority of MSC-SEK is validated by studying the creep-fatigue reliability evaluation of a typical aeroengine turbine disk. The current efforts open up an effective way to achieve high-accuracy and high-efficiency engineering creep-fatigue reliability evaluation.