文献类型：期刊文献

英文题名：Improved particle filter for fatigue crack propagation prediction using SH0 wave online monitoring

作者：Zhang, Yang[1];Jia, Jiuhong[1];Wang, Mingyuan[1];Gu, Yiqing[1];Tu, Shan-Tung[1]

机构：[1]East China Univ Sci & Technol, Key Lab Pressure Syst & Safety, Minist Educ, Shanghai 200237, Peoples R China

年份：2025

卷号：67

期号：9

起止页码：539

外文期刊名：INSIGHT

收录：;EI(收录号：20253919239452);WOS:【SCI-EXPANDED(收录号:WOS:001583343600001)】；

基金：The studies were funded by Innovative Groups Project (Grant Number 52321002) and Xingliao Talent Programme with the Soil Transplantation Project (XLYC2204030) .

语种：英文

外文关键词：SH0 wave monitoring; improved particle filter; hydrogenation reactor; fatigue crack propagation prediction; damage index

摘要：Accurately predicting crack propagation is fundamental in ensuring the safe operation of pressure equipment. Using particle filtering (PF) to predict crack propagation is an effective means, but classical particle filtering may exhibit insufficient particle diversity at the later stage of crack propagation, which makes it difficult to meet the requirements of high accuracy. To address this problem, in the present research, three improved crack propagation prediction models are established using three improved PF algorithms in combination with non-dispersive fundamental shear horizontal (SH0) waves. These three improved algorithms are classical algorithms to improve the prediction accuracy from three different perspectives and have been proved to be effective and reliable in prediction by several researchers. The three improved PF algorithms are auxiliary particle filter (APF), regularised particle filter (RPF) and Kullback-Leibler divergence (KLD) resampling particle filter. To verify the predictive effectiveness of the three improved algorithms in crack propagation prediction, a fatigue experiment is conducted in this study. The experiment results indicate that all three improved prediction models alleviate the particle diversity scarcity problem, leading to improved accuracy and reliability of the predictions. Among them, the prediction model with APF and SH0 wave monitoring has the smallest root mean square error in the prediction results. The prediction model with RPF and SH0 wave monitoring has the best a posteriori predictions estimated for the later crack propagation stages. The prediction model with KLD and SH0 wave monitoring has the shortest computing prediction time for a constant number of particles.