基于代理模型的广布疲劳损伤平均行为快速确定方法  

Rapid Determination of Average Behavior of Widespread Fatigue Damage Based on Proxy Model

作　　者：潘微 郑旭东 李杰 李彦 谈嘉好 陈昶 PAN Wei;ZHENG Xudong;LI Jie

机构地区：[1]航空工业成都飞机工业(集团)有限责任公司,成都610092

出　　处：《科技创新与应用》2023年第35期7-10,共4页Technology Innovation and Application

基　　金：2022年国家重点研发计划“高性能制造技术与重大装备”重点专项(2022YFB3404100)。

摘　　要：针对广布疲劳损伤平均行为预测过程复杂、消耗大量时间的问题,建立广布疲劳损伤概率萌生模型和裂纹扩展模型,在裂纹扩展过程中通过ABAQUS二次开发求解出七孔板孔边裂纹前缘的应力强度因子,应用蒙特卡罗法得到广布疲劳损伤平均行为,通过机器学习建立广布疲劳损伤仿真程序训练集,构建应力水平、裂纹随机萌生、裂纹随机扩展3个参数为设计变量,分别建立多项式响应面模型和高斯过程回归模型训练。结果表明,广布疲劳损伤平均行为与试验数据相比误差在6%之内,多项式响应面模型和高斯过程回归模型相比较,高斯过程回归模型精度更高,误差不超过10%,与传统方法相比可以节约90%的计算时间成本,所构建的高斯过程回归模型可以精确地确定广布疲劳损伤平均行为。Aiming at the problem that the prediction process of the average behavior of widespread fatigue damage is complex and takes a lot of time,the probabilistic initiation model of widespread fatigue damage and the model of crack propagation are established.In the process of crack propagation,the stress intensity factor at the leading edge of hole edge crack in seven-hole plate is solved by ABAQUS secondary development,the average behavior of widespread fatigue damage is obtained by Monte Carlo method,and the training set of widespread fatigue damage simulation program is established by machine learning.The stress level,random crack initiation parameters and random crack propagation are constructed as design variables,and the polynomial response surface model and Gaussian process regression model are established respectively.The results show that the error of the average behavior of widespread fatigue damage is less than 6%compared with the experimental data.Compared with the polynomial response surface proxy model and the Gaussian process regression proxy model,the accuracy of the Gaussian process regression model is higher,and the error is less than 10%.Compared with the traditional method,the calculation time cost can be saved by 90%.The Gaussian process regression model can accurately determine the average behavior of widespread fatigue damage.

关 键 词：广布疲劳损伤 概率分布 代理模型 高斯过程回归 蒙特卡罗法 

分 类 号：V215.5[航空宇航科学与技术—航空宇航推进理论与工程]