摘要
针对涡流激励热成像检测在精度与速度方面存在的短板,对快速区域卷积神经网络算法(Faster R-CNN)进行了改进,将Inception-V4结构引入其中作为特征提取网络,提高缺陷识别的精度与速度.通过实验获取不同类型缺陷数据,导入卷积神经网络进行训练与测试,并与常见VGG和ResNet网络所处理的结果进行对比.研究结果表明,改进后的网络算法可以实现缺陷识别的需求,在识别精度和识别速度方面均优于传统算法.
For eddy current excitation thermal imaging detection,the faster region convolutional neural network(Faster R-CNN)algorithm is improved,and the Inception-V4 structure is introduced as a feature extraction network to improve the accuracy and speed of defect recognition.Experimental methods are used to obtain different types of defect data,and convolutional neural networks are imported for training and testing,and the results are compared with the results processed by common VGG and ResNet networks.The results show that the model proposed in this paper can meet the requirements of defect recognition,and is better than the traditional algorithm in terms of recognition accuracy and recognition speed.
作者
李建宇
程姣姣
朱和明
郑文培
LI Jianyu;CHENG Jiaojiao;ZHU Heming;ZHENG Wenpei(State Key Laboratory of Shale Oil and Gas Enrichment Mechanisms and Effective Development,Beijing 102206,China;SINOPEC Research Institute of Petroleum Engineering Co.,Ltd.,Beijing 102206,China;Key Laboratory of Oil and Gas Safety and Emergency Technology,Ministry of Emergency,Beijing 102249,China)
出处
《河北大学学报(自然科学版)》
北大核心
2025年第6期653-660,共8页
Journal of Hebei University(Natural Science Edition)
基金
国家自然科学基金项目(51404283)。
关键词
无损检测
设备维护
故障诊断
缺陷识别
non-destructive testing
equipment maintenance
fault diagnosis
defect identification
