基于1D-CNN-SVM的钢丝绳损伤识别方法  

Wire rope damage identification method based on 1D-CNN-SVM

作　　者：任建浩 陈实 薛家杰 吴非 孙燕华[1] REN Jianhao;CHEN Shi;XUE Jiajie;WU Fei;SUN Yanhua(School of Mechanical Science and Engineering,Huazhong University of Science and Technology,Wuhan 430074,China;CNOOC Environmental Protection Services(Tianjin)Co.,Ltd.,Tianjin 300450,China)

机构地区：[1]华中科技大学机械科学与工程学院,武汉430074 [2]中海石油环保服务(天津)有限公司,天津300450

出　　处：《无损检测》2024年第6期24-29,共6页Nondestructive Testing

基　　金：国家自然科学基金(52275532);交通运输部纵向项目(SXHXGZ-2021-2);国家重点研发纵向项目(2021YFF0501000)。

摘　　要：为实现对钢丝绳损伤的高效、精准识别,提出一种基于1D-CNN-SVM的钢丝绳损伤识别模型。使用一维卷积神经网络对损伤的漏磁检测信号进行特征提取,然后输入到支持向量机中进行缺陷分类,通过将不同工况速度下的数据集代入该模型,检验所提模型的缺陷识别能力。试验结果表明,相较于1D-CNN,1D-CNN-ELM,1D-CNN-LTSM等模型,所提模型的准确性和可靠性更高,对各类损伤的识别准确率均不小于97%,体现出较强的泛化能力。To achieve efficient and accurate identification of wire rope damage,a wire rope damage identification model based on 1D-CNN-SVM was proposed.A one-dimensional convolutional neural network was used to extract features from the magnetic flux leakage detection signal of damage,and then extracted features were input into a support vector machine for defect classification.By substituting datasets from different operating speeds into the model,the defect recognition ability of the proposed model was tested.The experimental results showed that compared to models such as 1D-CNN,1D-CNN-ELM,and 1D-CNN-LTSM,the proposed model had higher accuracy and reliability,with an accuracy rate of no less than 97%for identifying various types of damage,demonstrating strong generalization ability.

关 键 词：一维卷积神经网络 支持向量机 漏磁检测 损伤识别 

分 类 号：TG115.28[金属学及工艺—物理冶金]