压力容器2025，Vol.42Issue(9)：71-79,9.DOI:10.3969/j.issn.1001-4837.2025.09.009
基于Co-RF-SVM协同训练的半监督算法弱磁信号信噪辨识
A semi-supervised algorithm based on Co-RF-SVM cooperative training for signal and noise identification of weak magnetic signals
摘要
Abstract
In order to solve the problem that weak magnetic detection of pipeline is easily disturbed by noise,a signal and noise identification model based on semi-supervised classification cooperative training algorithm(Co-RF-SVM)for pipeline magnetic memory detection of weak magnetic signals is established.By prefabricating L245N steel tube specimens with valves and flanges,the influence of magnetic interference sources such as pipeline structures and side-by-side pipes on magnetic memory signals was studied.The results show that the gradient value of each magnetic field component under the influence of the valve and flange has a sudden change region,and the side-by-side pipes will deflect or weaken the magnetic memory signal.The peak value of the synthetic magnetic field intensity of the magnetic memory signal gradually increases with the increase of pipeline distance.When the distance of the pipeline is about 105 mm,it has the greatest influence on the magnetic memory signal.For the abnormal signal regions located,six parameters were extracted as characteristic values,including magnetic field intensity change rate,gradient limit state coefficient,approximate entropy,sample entropy,fuzzy entropy and singular spectrum entropy.Defects,valves and flanges were used as labels to construct data sets.A semi-supervised cooperative training algorithm based on random forest and support vector machine was introduced to establish a signal and noise identification classification model for the sudden change area of pipeline magnetic memory signal.The model verification results show that the classification accuracy of defects,valves and flanges are 86%,90%and 94%,respectively.This study provides a new tool for solving the problem of noise interference in pipeline weak magnetic detection.关键词
油气管道/磁记忆检测/降噪/半监督分类算法Key words
oil and gas pipeline/magnetic memory detection/noise reduction/semi-supervised classification algorithm分类
机械制造引用本文复制引用
吴思炫,邢海燕,张海,段成凯,赵力伟,蔡智会,袁钟秀,王朝玺..基于Co-RF-SVM协同训练的半监督算法弱磁信号信噪辨识[J].压力容器,2025,42(9):71-79,9.基金项目
黑龙江省自然科学基金联合引导项目(LH2024E012) （LH2024E012）
国家级大学生创新创业训练计划项目(202310220009) （202310220009）
温州市市场监督管理局科研计划项目(2024011) （2024011）
