2026年“核理安邦”联合博士生学术论坛

AM-MCMC与RPF耦合的概率断裂力学方法及CRDM飞射物风险量化研究

Not scheduled

1h

地址：清华大学校内

海报展示 海报展示 海报展示

Speaker

张 贺 (哈尔滨工程大学)

Abstract

In accordance with the HAD102/04-2019 Nuclear Safety Guideline, a full-life failure probability below 10−7 is required for a nuclear power plant Control Rod Drive Mechanism (CRDM) to qualify as an incredible projectile. Conventional static Probabilistic Fracture Mechanics (PFM) methods are inherently overly conservative and incapable of assimilating in-service inspection data. Taking the HPR1000 ML-B+ type CRDM as the research object, this paper assesses the dominant failure mode of Stress Corrosion Cracking (SCC). A semi-elliptical crack defect model incorporating thermal aging effects and nonlinear residual stresses is constructed, excluding the crack initiation phase. A high-order SCC growth model and a surrogate model with relative error ≤ 5% are developed. A quasi-dynamic Bayesian framework coupling Adaptive Markov Chain Monte Carlo (AM-MCMC) and Regularized Particle Filter (RPF) is proposed to enable unbiased parameter identification and dynamic failure probability updating. Monte Carlo simulations show that the failure probabilities of all critical sections and welds are consistently on the order of 10−8, well below the safety threshold. The full-life evolution of failure probabilities is stable without abrupt increases, and initial defect depth is identified as the dominant controlling factor via sensitivity analysis. Noisy Observing System Simulation Experiment (OSSE) tests verify that the tracking accuracy and convergence performance of the algorithm meet engineering requirements. The proposed method overcomes the limitations of conventional PFM approaches, validates the incredible projectile compliance of the CRDM, and provides a quantitative basis for full-life safety assessment of key nuclear power components.

摘要

依据 HAD102/04-2019 核安全导则，核电厂控制棒驱动机构（CRDM）全寿期失效概率低于10⁻⁷方可判定为不可信飞射物。针对传统静态概率断裂力学（PFM）方法过度保守、无法同化在役数据的缺陷，本文以华龙一号ML-B+型CRDM为对象，聚焦应力腐蚀开裂（SCC）失效模式开展评估。构建含热老化效应与非线性残余应力的半椭圆缺陷模型，剔除裂纹萌生阶段，搭建高阶SCC扩展模型与误差≤5% 的代理模型；提出自适应马尔可夫链蒙特卡洛（AM-MCMC）与正则化粒子滤波（RPF）耦合的贝叶斯准动态框架，实现参数无偏识别与失效概率动态更新。蒙特卡洛仿真表明，关键截面与焊缝失效概率均稳定处于 10⁻⁸量级，远低于安全阈值；全寿期演化平稳，初始缺陷深度为核心主控因素。OSSE含噪测试验证算法跟踪精度与收敛性满足工程要求。本方法突破传统PFM局限，验证了CRDM的不可信飞射物合规性，为核电关键部件全寿期安全评估提供支撑。