Elbow pipe components are frequently subjected to complicated thermo-mechanical load combinations cyclically in nuclear engineering, facing failures related to cyclic plastic responses, including low cycle fatigue (LCF) and ratcheting. To deal with the risk management of important pipelines, the probabilistic LCF and ratcheting assessments are indispensable to evaluate the reliability of such components considering the uncertain operating parameters. In this study, the new probabilistic Linear Matching Method (pLMM) framework is proposed to address the probabilistic structural integrity assessment, quantitatively predicting the statistical distribution of LCF life and ratchet limit by the surrogate model with the novel Linear Matching Method-driven neural network (LDNN). With the numerical investigations on the elbow pipe structure presented, the probabilistic assessment boundaries and reliability-centred evaluation diagrams in terms of LCF life and ratchet limit are established respectively, which are beneficial to get rid of the conservativeness of the traditional design schemes with safety factor.

弯管部件在核工程中经常承受复杂的热机械载荷组合，周期性地承受，面临与周期性塑性响应相关的故障，包括低周疲劳 (LCF) 和棘轮效应。为了应对重要管道的风险管理，概率LCF和棘轮评估在考虑不确定运行参数的情况下评估此类组件的可靠性是必不可少的。在这项研究中，提出了新的概率线性匹配方法 (pLMM) 框架来解决概率结构完整性评估问题，通过具有新型线性匹配方法驱动的神经网络的代理模型定量预测 LCF 寿命和棘轮极限的统计分布（低密度神经网络）。通过对弯管结构的数值研究，