A data-driven roadmap for creep-fatigue reliability assessment and its implementation in low-pressure turbine disk at elevated temperatures
本文提出了一种基于数据驱动的蠕变疲劳可靠性评估路线图，以低压涡轮盘为例，通过优化机器学习代理模型提高计算效率，并采用概率损伤交互图方法进行保守的可靠性评估。
High-reliability life design process not only can ensure system safety in service, but also can provide scientific life management during the maintenance period. The objective of the present work is to develop a roadmap for creep-fatigue reliability assessment. Material-level data accumulations and theoretical foundations of creep-fatigue including creep-fatigue constitutive and multi-axial damage models are introduced. Afterwards, a low-pressure turbine disk under a typical creep-fatigue load waveform is applied as a case study to demonstrate how to perform creep-fatigue reliability assessment by using this roadmap in practice. Precise weakness hotspots are identified at the mortise joint of turbine disk. Based on hotspot-based strategy, it is found that the surrogate model assisted by an optimized machine learning method enhances significantly the computational efficiency. Accordingly, the probabilistic creep-fatigue life with considering multi-sources uncertainty obeys lognormal distributions. In the aspect of failure probability analysis, the current probabilistic damage interaction diagram method with creep-fatigue interaction gives conservative reliability assessments and excellent universality as compared to traditional ones mainly used in the low cycle fatigue field. Last but not least, joint failure evaluation of the turbine disk is discussed to comprehensively consider potential failure occurrence in an averaged hot region instead of a single hotspot.