卷积神经网络在燃气管道故障诊断中的应用

详细信息    查看全文 | 推荐本文 |

英文篇名：Research on Convolution Neural Network in Gas Pipeline Fault Diagnosis 作者：王新颖 ; 杨泰旺 ; 宋兴帅 ; 陈海群 英文作者：WANG Xinying;YANG Taiwang;SONG Xingshuai;CHENG Haiqun;School of Environmental & Safety Engineering,Changzhou University; 关键词：燃气管道 ; 故障诊断 ; 卷积神经网络 ; 数据参数分析 英文关键词：gas pipeline;;fault diagnosis;;convolution neural network;;data parameter analysis 中文刊名：工业安全与环保 英文刊名：Industrial Safety and Environmental Protection 机构：常州大学环境与安全工程学院; 出版日期：2019-02-10 出版单位：工业安全与环保 年：2019 期：02 基金：常州市国际科技合作项目 语种：中文; 页：40-44+72 页数：6 CN：42-1640/X ISSN：1001-425X 分类号：TP183;TU996.7

摘要

为了提高燃气管道泄漏故障的诊断能力,判断燃气管道故障泄漏的类型,将深度学习神经网络应用到燃气管道故障诊断领域,提出了一种基于卷积神经网络与softmax分类器的燃气管道故障诊断技术。选取燃气管道故障中的9种特征参数作为模型的原始输入量,经过卷积神经网络的特征提取、参数重构、soft-max分类,最终达到诊断的效果。将得到的卷积神经网络模型应用在实验室的燃气管道故障泄漏检测系统中,结果表明,这种模型在燃气管道故障识别中的准确率稳定在92. 54%,与BP神经网络相比该方法有明显的准确率和稳定性。
        In order to improve the diagnostic capability of gas pipeline leakage failure and predict the risk of gas pipeline fault leakage,the depth learning neural network is applied in the field of gas pipeline fault diagnosis,and a gas pipeline fault diagnosis technology based on convolution neural network and softmax classifier is proposed. The nine characteristic parameters in the gas pipeline failure are selected as the original input of the model,and the feature extraction,parameter reconstruction and softmax classification of the convolution neural network are used to achieve the effect of diagnosis. The obtained convolution cloud neural network model is applied in the laboratory fault detection system of gas pipeline fault and the results show that the accuracy of this model is 92. 54% in the gas pipeline fault identification,and the method has obvious accuracy and stability compared with other methods such as BP.

引文

[1]叶迎春,张来斌,王金江.输气管道泄漏声源特性及其变化规律[J].天然气工业,2016,36(8):124-131.
    [2]禹建丽,卞帅.基于BP神经网络的变压器故障诊断模型[J].系统仿真学报,2014(6):1343-1349.
    [3]宋志杰,王健.模糊聚类和LM算法改进BP神经网络的变压器故障诊断[J].高压电器,2013(5):54-59.
    [4]符杨,江玉蓉,崔椿洪,等.基于模糊数学和概率论的变压器故障诊断[J].高电压技术2008(5):1040-1044.
    [5]王新颖,江志伟,于永亮,等.多信息融合的城市燃气管道泄漏诊断技术研究[J].中国安全科学学报,2014,24(6):165-171.
    [6]Ai Lingmei,Wang Jue,Wang Xuelian. Multi-features fusion Diagnos is of tremor based on artificial neural network and D-S evidence theory[J]. Signal Processing,2008,88:2927-2934.
    [7]G E Hinton,S Osindero and Y W. A fast learning algorithm for deep belief nets[J]. In Neural Computation,2006,18(7):1527-1554.
    [8]刘建伟,刘媛,罗雄麟.深度学习研究进展[J].计算机应用研究,2014,31(7):1921-1930.
    [9]张晴晴,刘勇,潘接林,等.基于卷积神经网络的连续语音识别[J].工程科学学报,2015(9):1212-1217.