基于IPSO-PF算法的疲劳裂纹扩展预测  

Fatigue crack growth prediction based on IPSO-PF algorithm

作　　者：靳婷 王晓磊 刘宇[1] 袁建明[1] JIN Ting;WANG Xiaolei;LIU Yu;YUAN Jianming(Key Laboratory of Port Cargo Handling Technology Ministry of Communications,School of Transportation and Logistics,Wuhan University of Technology,Wuhan 430063,China;Installation Engineering Co.,Ltd.,China Communications First Harbor Engineering,Tianjin 300457,China)

机构地区：[1]武汉理工大学交通与物流工程学院,港口装卸技术交通运输行业重点实验室,武汉430063 [2]中交一航局安装工程有限公司,天津300457

出　　处：《机械强度》2025年第4期47-53,共7页Journal of Mechanical Strength

基　　金：国家重点研发计划项目(2022YFB2602302)。

摘　　要：传统Paris公式预测裂纹扩展时忽略了裂纹扩展过程中各种不确定因素的影响,导致预测的裂纹扩展过程与真实的裂纹扩展过程相差较大。为提高疲劳裂纹扩展预测的精度,提出了一种基于改进粒子群优化粒子滤波(Improved Particle Swarm Optimization-Particle Filtering,IPSO-PF)算法的疲劳裂纹扩展预测方法。首先,在粒子滤波(Particle Filtering,PF)算法的框架上,利用粒子群优化(Particle Swarm Optimization,PSO)算法对基于观测信息更新后的部分粒子进行优化,保持大权值的粒子状态不变,将小权值的粒子趋向于高似然区域,设计了IPSO-PF算法;然后,将IPSO-PF算法与Paris公式结合,构建了基于Paris公式和IPSO-PF算法的疲劳裂纹扩展预测模型;最后,使用公开的2024-T351铝合金数据集对该模型的有效性进行了验证。结果表明,与传统PF算法相比,IPSO-PF算法能够提高粒子的多样性,使用IPSO-PF算法构建的裂纹扩展预测模型的预测误差为2.6%,优于基于PF算法的9.2%。The traditional Paris formula ignores the influence of various uncertain factors in the crack growth process,which leads to a big difference between the predicted crack growth process and the real crack growth process.In order to improve the prediction accuracy of fatigue crack growth,a fatigue crack growth prediction method based on the improved particle swarm optimization particle filtering(IPSO-PF) algorithm was proposed.Firstly,based on the framework of the particle filtering(PF) algorithm,the particle swarm optimization(PSO) algorithm was used to optimize some particles based on the updated observation information,keeping the state of particles with large weights unchanged,and particles with small weights tend to high likelihood region,and IPSO-PF algorithm was designed.Then,combining IPSO-PF algorithm with Paris formula,a fatigue crack growth prediction model based on Paris formula and IPSO-PF algorithm was constructed.Finally,the validity of the model was verified by using the open 2024-T351 aluminum alloy data set.The results show that compared with the traditional PF algorithm,IPSO-PF algorithm can improve the diversity of particles.The prediction error of the crack growth prediction model based on IPSO-PF algorithm is 2.6%,which is better than 9.2% based on PF algorithm.

关 键 词：疲劳裂纹 裂纹扩展预测 粒子滤波 粒子群优化 算法优化 

分 类 号：O346.2[理学—固体力学]