基于改进DRSN-CW的气体泄漏检测技术  

Gas leakage detection technology based on improved DRSN-CW

作　　者：李德望 周瑶 LI Dewang;ZHOU Yao(Lijiang Power Supply Bureau of Yunnan Power Grid Co.,Ltd.,Lijiang 674100,China)

机构地区：[1]云南电网有限责任公司丽江供电局,云南丽江674100

出　　处：《中国高新科技》2024年第22期140-142,共3页

摘　　要：为了解决环保气体泄漏检测问题,文章提出了一种基于通道阈值深度残差网络(DRSN-CW)的改进方法,以提高GPLA-12数据集的检测精度。在该方法中,所有卷积层都设计了更大且不等的卷积核大小,以便在提取故障特征的过程中扩展感受野。此外,考虑到环保气体泄漏数据集通常含有大量的环境噪声,DRSN-CW的软阈值模块与设计的核大小相结合,降低了噪声对环保气体泄漏检测精度的影响。实验结果表明,基于改进DRSN-CW气体的泄漏检测技术优于目前常用的技术。In order to solve the problem of environmental gas leakage detection,this paper proposes an improved method based on channel threshold deep residual network(DRSN-CW)to improve the detection accuracy of the GPLA-12 dataset.In this method,all convolutional layers are designed with larger and unequal convolutional kernel sizes to expand the receptive field during the process of extracting fault features.In addition,considering that environmental gas leakage datasets typically contain a large amount of environmental noise,the combination of DRSN-CW’s soft threshold module and the designed kernel size reduces the impact of noise on the accuracy of environmental gas leakage detection.The experimental results show that the leakage detection technology based on improved DRSN-CW gas is superior to the commonly used technologies at present.

关 键 词：安全 泄漏检测 基于信道的深度残余收缩网络 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]