摘要
燃气轮机滑油管状态红外监测是开展滑油管故障诊断的基础,开展燃气轮机滑油管红外监测与识别研究对及时掌握燃气轮机的工作状态有重要意义。本文依据在不同工况下实验采集得到的某型燃气轮机滑油管的红外热像图,采用Faster RCNN算法对得到的图像进行训练和识别,结果显示该算法能够精确识别燃气轮机的滑油管等不同部位。对比4种迁移网络的训练和测试结果,发现在有背景干扰情况下Resnet50迁移网络对滑油管部件的识精度不高,而在兼顾网络检测时间和目标识别精度的情况下Vgg16迁移网络最优。
Infrared monitoring of gas turbine oil pipe condition is the basis for fault diagnosis of oil pipe and it is of great significance to carry out infrared monitoring and identification research on gas turbine oil pipe to master the working state of gas turbine in time.In this paper,the Faster RCNN algorithm is used to train and recognize the images based on the infrared thermal image of a certain type of gas turbine oil pipe collected under different working conditions.The results show that the algorithm can accurately identify different parts of gas turbine such as oil pipe.Comparing the training and testing results of the four migration networks,it is found that Resnet50 migration network has lower identification accuracy for the tubing components under background interference,while the Vgg16 migration network is the best in terms of network detection time and target recognition accuracy.
作者
刘寅
夏舸
王强
谢志辉
杨立
LIU Yin;XIA Ge;WANG Qiang;XIE Zhi-hui;YANG Li(College of Power Engineering,Naval University of Engineering,Wuhan 430033,China)
出处
《激光与红外》
CAS
CSCD
北大核心
2023年第4期544-550,共7页
Laser & Infrared
基金
航空发动机及燃气轮机重大专项项目(No.J2019-Ⅳ-0021-0089)资助。
