集总经验模态分解法提取光纤振动入侵信号的研究

Optical fiber sensing technology has brought new hope to the pipeline safety precautions in the development and application of world oil and gas storage and transportation industry. It can carry out early detection, warning and location of the acts before the detected damage and destruction of the threats, so pipeline operators have plenty of time to stop the destruction of behavior and reduce the loss changing the current passive situation of pipeline transportation. The research topic introduces ensemble empirical mode decomposition(EEMD) into.analysis of optical fiber vibration signal, extracting fault model and early warning. Firstly, vibration signal containing noise is decomposed into a number of intrinsic mode function(IMF) components whose scale is gradually increasing after EEMD. Secondly, the IMFs are obtained from the noisy signal by threshold denoising method based on the signal and noise energy distribution characteristics. Thirdly, Hilbert energy spectrum obtaining sub-band energy(SE), DSE, SECC, DSECC is got by applying the Hilbert transform to IMFs, after the masking effect to divide.vibration signal into several frequency band. Finally, vibration feature vector composited of characteristics of each band and normalized short-term energy by the Teager operator is inputted into the fuzzy least squares support vector machine to recognize different fault mode. Vibrating feature based on EEMD, the Hilbert energy spectrum and Teager energy provides new ideas for fault diagnosis with noisy environments and remote detection. When mining or illegal intrusion events exists in the vicinity of the pipeline, the optical fiber vibration can recognize.fault and make alarm, with accurate positioning vibration position, as far as possible to retrieve the economic loss.

光纤传感技术在世界油气储运界的开发和应用给管道安全防范带来了新的希望，它能够在对管道造成破坏之前，预先检测出破坏行为的威胁并及时告警，给管道运营商留有足够的时间来制止破坏行为，扭转管道运输业当前被动局面，减少损失。课题研究集总经验模态分解方法分析光纤振动信号，提取故障模式，识别并预警。首先采用集总经验模态分解包含噪声的振动信号，筛选本征模态函数，利用软阈值方法甄别噪声和振动频谱，实现振动信号的预处理。进一步对各组本征模态函数进行Hilbert变换，利用掩蔽效应将信号所在频段划分若干频带，获得Hilbert能量谱，计算子带能量及其差分、和Teager复合能量、过零率等一起组成振动特征向量，输入模糊最小二乘支持向量机进行判断，从而对大噪声、长距离的振动状态进行识别，当在管线附近有挖掘或者非法入侵事件，诊断并告警，精确定位振动位置，尽可能的挽回经济损失。

输油管道是石油的主要运输途径之一，主要敷设的区域大多在偏僻的地区，容易受到人类活动、动植物运动、灾害性天气、地质灾难等非法入侵的危害，输油管道的日常维护和安全保障难度较高。输油管道周界防范系统的研究和设计是国内外的研究热点。. 本项目设计了埋地光纤型输油管道周界防范系统，在充分考虑入侵信号的环境等噪声干扰基础上，深入分析了光纤振动信号的时频特性，提出了集总经验模态分解振动信号的方法，避免了经验模态分解算法可能存在的模态混叠现象。并利用软阈值滤除噪声获取本征模态函数。深入研究了频谱质心等特征，并将其引入入侵信号诊断算法中，将剔除噪声的本征模态函数分量组合形成重构信号，进行希尔伯特变换获取能量谱，提取特征参数，构造特征向量。在深入研究识别网络的基础上，提出了最小二乘支持向量机状态分类器，将特征向量导入最小二乘支持向量机进行学习和分类，从而诊断入侵信号，并进一步确定非法入侵的区域和发出报警。利用分析光纤型输油管道周界防范系统采集到的现场数据进行分析，与特征模态奇异值分解等方法进行对比分析，论证了项目所提方法的有效性。. 本项目所提出的入侵诊断和分析方法，可以为输油管线、设备运行等异常状态检测、故障诊断提供理论依据和技术支持，有利于提高我国周界防范系统的研究水平，有较强的学术和应用价值。