基于TPE-XGBoost算法的镍基600合金应力腐蚀裂纹扩展速率预测模型

Prediction of Stress Corrosion Crack Growth Rate of Ni-base Alloy 600 Based on TPE-XGBoost Algorithm

运行经验表明,应力腐蚀开裂(SCC)是镍基600合金在压水堆核电站一回路高温高压水环境中的主要失效形式。针对镍基600合金SCC影响因素多、机理复杂,现有预测模型应用性不高的问题,利用TPE-XGBoost算法,通过机器学习挖掘应力强度因子、温度、屈服强度、溶解氢含量、裂纹扩展方向、载荷类型、热处理工艺等影响因素与裂纹扩展速率之间的关系,建立反映多维数据关联关系的非参数镍基600合金应力腐蚀裂纹扩展速率预测模型。结果表明,TPE-XGBoost算法可以实现高维度数据集超参数快速优化,且有效避免优化结果陷入局部最优解,使得预测模型具有良好的泛化能力,将应用于压水堆核电站镍基600合金部件反应堆冷却剂条件应力腐蚀裂纹扩展工程预测。

Stress corrosion cracking(SCC) endangers structural integrity of the nickel-base alloy 600 components widely used in the water environment of high temperature and high pr essure in pressurized water reactors(PWRs). Due to the complexity of the interweaving influences, the existing prediction models developed for SCC are limited for engineering assessment by accuracy. In this study, a non-algebraic model with multi-dimensional data associations was developed for predicting the SCC growth rate of the Ni-base alloy 600, which utilized the TPE-XGBoost machine learning algorithm to describe the correlation between the multiple characteristic parameters including stress intensity factor, temperature, yield strength, dissolved hydrogen content, crack propagation direction, load type, heat treatment process and SCC growth rate. It is found that the TPE-XGBoost algorithm could achieve rapid global optimization of multi-dimensional data sets rather than the local optimal values. The obtained SCC model with sound generalization ability demonstrates potential engineering application on SCC growth rate prediction of Ni-base alloy 600 components in PWRs.