基于音波的输气管道泄漏检测技术研究

发布时间：2018-07-16 21:02

【摘要】：在长输气管道的运行中,由于自然腐蚀、老化、人为损坏等原因,管道泄漏情况时有发生,天然气具有易燃易爆的特点,如果泄漏情况不能被及时发现,就可能给社会和环境带来重大损失,甚至造成人员伤亡。因此对输气管道泄漏检测技术的研究就变得尤为重要。本文依托高压输气管道实验系统,对基于音波法的泄漏检测技术进行了研究。为了降低采集到的音波信号的噪声,采用小波变换的方法进行去噪,确定了去噪的最优小波基、分解层数、阈值处理方法等各项参数。通过对音波信号时域、频域、小波域的分析,提取出可用于泄漏判断的特征量。选取合适的特征量作为神经网络的输入,以是否发生泄漏为输出建立了BP神经网络,实现泄漏的准确判断。最后完成了基于LabVIEW的泄漏检测系统的设计。本文通过试验研究与理论分析,得出以下主要结论:(1)系统采集的音波信号属于非平稳信号,适用于小波变换法去噪。(2)确定了小波阈值法去噪参数,最优小波基为’sym2’小波基,11层分解,阈值选取规则为试探法的Stein无偏风险阈值(heursure)规则,软阈值处理方法。(3)利用连续小波实现信号的时频分析,最优小波基为‘mexh’小波。(4)均值、方差、斜率均值、自相关、频谱、功率谱、小波能量等可以作为泄漏判断的特征量。(5)建立了多参量的泄漏诊断系统,选取均值、方差、自相关、功率谱密度、小波包频带1能量这五个特征量,处理后组成特征向量,建立基于BP神经网络的多参量泄漏诊断系统,可以实现准确诊断。(6)设计了基于LabVIEW的泄漏检测系统,利用LabVIEW实现了小波去噪、特征提取、神经网络诊断、互相关法排除了工况干扰,为算法在工程上的应用提供了参考。
[Abstract]:In the operation of long gas transmission pipeline, due to natural corrosion, aging, artificial damage and other reasons, pipeline leakage occurs, natural gas has the characteristics of flammable and explosive. If the leakage situation can not be found in time, it may bring great loss to the society and the environment and even cause casualties. Therefore, the leakage detection technology of gas pipeline is used. In this paper, based on the experimental system of high pressure gas transmission pipeline, the leakage detection technology based on sound wave method is studied. In order to reduce the noise of the acoustic wave signal, the method of wavelet transform is used to denoise, and the parameters of the optimal wavelet base, the decomposition layer number, the threshold processing method and so on are determined. After the analysis of the time domain, frequency domain and wavelet domain of the sound wave signal, the characteristic quantity can be extracted for leakage judgment. The appropriate characteristic amount is selected as the input of the neural network, and the BP neural network is established to determine the leakage and the leakage is accurately judged. Finally, the design of the leakage detection system based on LabVIEW is completed. This paper passes through this paper. The main conclusions of the experimental and theoretical analysis are as follows: (1) the sound wave signals collected by the system are non-stationary signals, and are suitable for wavelet transform denoising. (2) the wavelet threshold method is used to determine the denoising parameters, the optimal wavelet base is "sym2" small wave base, the 11 layer decomposition, the threshold selection rule is the Stein unbiased risk threshold (heursure) rule of the test method, Soft threshold processing method. (3) using continuous wavelet to realize signal time-frequency analysis, the optimal wavelet base is' mexh 'wavelet. (4) mean, variance, slope mean, autocorrelation, spectrum, power spectrum, wavelet energy and so on. (5) a multi parameter leakage diagnosis system is established, and the mean, variance, autocorrelation, power spectrum are selected. Density, wavelet packet frequency band 1 energy of these five characteristics, after processing the composition of the feature vector, the establishment of a multi parameter leakage diagnosis system based on BP neural network, can achieve accurate diagnosis. (6) the design of LabVIEW based leakage detection system, the use of LabVIEW to achieve wavelet denoising, feature extraction, neural network diagnosis, mutual clearance method to exclude working conditions Interference provides a reference for the application of algorithm in engineering.
【学位授予单位】：中国石油大学(华东)
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE973.6

【参考文献】