摘 要:现代发电厂中蒸汽管道是生产的生命线,蒸汽泄漏检测关乎着设备正常运转。针对电厂生产经营中蒸汽管道可能出现的蒸汽泄漏情况,解决蒸汽泄漏检测的实时性与准确性问题,应用 YOLOv5s 目标检测算法进行研究。通过采集不同场景条件下的数据集,并进行训练并调优,最终实验指标显示,训练模型具有良好的测试效果,针对实际复杂场景下的蒸汽泄漏情况具有较强的泛化能力和检测能力,能够满足现有场景下电厂蒸汽泄漏检测的实时性和准确性要求。
关键词:电厂蒸汽泄漏检测;YOLOv5s;目标检测;深度学习
DOI:10.19850/j.cnki.2096-4706.2022.19.013
中图分类号:TP391 文献标识码:A 文章编号:2096-4706(2022)19-0052-06
Research on Power Plant Steam Leak Detection Based on YOLOv5s
YE Jianqing, WANG Yihe, MU Jiuzhuang, WU Bing, ZHU Feng
(Guoneng Shenwan Chizhou Power Generation Co., Ltd., Chizhou 247100, China)
Abstract: Steam pipeline is the lifeline of production in modern power plant, and steam leak detection is related to the normal operation of equipment. Aiming at the possible steam leakage situation of steam pipeline in power plant production and operation, in order to solve the problems of real-time and accuracy of steam leak detection, YOLOv5s object detection algorithm is applied to study. Through collecting data set under the conditions of different scenarios and training and tuning, finally the experimental indicators show that training model has good test effect and stronger generalization ability and detecting ability in view of the steam leak situation under the actual complex scenarios. And it can satisfy the real-time and accuracy requirements of power plant steam leak detection under the existing scenarios.
Keywords: power plant steam leak detection; YOLOv5s; object detection; deep learning
参考文献:
[1] RIVERS HK,SIKORA JG,SANKARAN SN.Detection of hydrogen leakage in a composite sandwich structure at cryogenic temperature [J].J Spacecr Rockets,2002,39(3):452-459.
[2] 马胤刚,张冠男,王明威,等 . 基于双目视觉的蒸汽泄露定位系统及方法:CN109854964A [P].2019-06-07.
[3] 龚其春,刘成良,王永红,等 . 新型气体泄漏超声检测系统的研究与设计 [J]. 液压与气动,2005(3):75-77.
[4] 王涛,覃鹤宏,赵苓,等 . 基于模糊核聚类图像分割的气体泄漏定位研究 [J]. 北京理工大学学报,2013,33(3):280-284.
[5] 李智华,易力 . 基于人工智能技术的电厂跑冒滴漏智能检测系统研究 [J]. 中国设备工程,2021(22):160-161.
[6] GIRSHICK R,DONAHUE J,DARRELL T. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation [C]//2014 IEEE Conference on Computer Vision and Pattern Recognition.Columbus:IEEE,2014:580-587.
[7] REDOM J,DIVVALA S,GIRSHICK S. You Only
Look Once:Unified,Real-Time Object Detection [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas:IEEE,2016:779-788.
[8] 涂沛驰,傅钰雯,熊宇璇,等 . 基于 YOLOv5 网络模型的火焰检测 [J]. 智能计算机与应用,2022,12(3):158-161.
[9] 代牮,赵旭,李连鹏,等 . 基于改进 YOLOv5 的复杂背景红外弱小目标检测算法 [J]. 红外技术,2022,44(5):504-512.
[10] 刘超阳,曲金帅,范菁,左金花,唐玉敏 . 基于改进YOLOv5 算法的车辆目标检测 [J/OL]. 云南民族大学学报(自然科学版 ):1-9[2022-09-21].http://kns.cnki.net/kcms/ detail/53.1192.N.20220421.0940.015.html.
[11] NEUBECK A,GOOL L V. Efficient Non-Maximum Suppression [C]//18th International Conference on Pattern Recognition (ICPR’06).Hong Kong:IEEE,2006:China.
[12] 彭道刚,刘薇薇,戚尔江,等 . 基于 CBAM-Res _UNet 电厂高压蒸汽泄漏检测研究 [J]. 电子测量与仪器学报,2021,35(12):206-214.
[13] 李昌夏,加文浩,黄政龙,等 . 基于 YOLOv5 的实时抽烟检测研究 [J]. 电脑知识与技术,2022,18(8):100-102.
[14] 李赞 . 基于深度学习的火灾检测研究与实现 [D]. 银川:宁夏大学,2021.
[15] 陈俊 . 基于 YOLOv3 算法的目标检测研究与实现 [D]. 成都:电子科技大学,2020.
[16] 代虎林 . 典型深度学习框架在单 GPU 环境中的模型训练性能分析 [D]. 武汉:华中科技大学,2020.
[17] 杨典,李小燕,刘培焱,等 . 基于 OpenCV 的变电站仪表识别方法研究 [J]. 自动化与仪表,2022,37(4):75-80.
作者简介:叶剑青(1980—),男,汉族,工程师,本科,研究方向:发电企业设备管理及安全管理。