小波分析在信号处理中的应用

在信号奇异点检测中,首先对信号进行多尺度小波分解。然后对高频部分进行重构,确定模极大值点位置,从而确定出奇异点位置。在例子中检测含有故障信息的电压信号,结果表明信号分解的细节部分均能清晰地显示出奇异点的准确位置。在信号消噪中,首先将信号进行多尺度小波分析。然后通过设置阈值的方法对高频子带进行处理,接着进行重构。在例子中对染噪的电压信号进行消噪处理,结果表明阈值的选取直接关系到消噪的质量。     

基于DOFVS的新型压力输水管道泄漏在线监测方法

压力输水管道因内部压力及外部使用环境腐蚀等因素经常造成爆管泄漏等问题,根据管道泄漏时会引起泄漏点周围管壁振动这一特点,利用基于相位敏感光时域反射仪技术的分布式光纤振动传感技术(DOFVS)提出了一种新型压力输水管道光纤在线泄漏监测方法,此方法利用普通单模通信光纤拾取泄漏点引起的管道振动信号并进行实时检测和定位。在室内测试环境下,该系统能够检测出DN90 cm×EN2 cm普通钢制压力输水管道在0.4 Mpa压力下,泄漏孔径为4 mm的泄漏;在室外测试环境下,该系统能够检测出DN200 cm×EN2 cm普通钢制压力输水管道在0.27 Mpa压力下,泄漏量大于11 L/s泄漏孔的泄漏。此外,采用多尺度小波分解去噪方法,对监测信号中的环境噪声信号进行滤除,并取得了良好的去噪效果。     

基于BWO优化VMD联合小波阈值的管道泄漏次声波去噪方法

管道泄漏次声波信号中的干扰噪声影响管道泄漏定位的精准度。提出了一种基于白鲸优化算法（BWO）优化变分模态分解（VMD）联合小波阈值（WT）的管道泄漏次声波去噪方法。针对VMD算法中分解层数K和惩罚因子α的取值对信号分解结果影响较大,利用白鲸优化算法（BWO）对VMD分解的两关键参数进行寻优,获得最优参数组合[K、α],并利用优化后的参数对次声波信号进行VMD分解,获得一系列本征模函数（IMF）分量。通过计算各IMF分量的相关系数来区分噪声IMF分量和有效IMF分量,引入一种改进的小波阈值函数对有效的IMF分量进行去噪处理,再重构去噪后各有效IMF分量,得到去噪后的管道泄漏次声波信号。通过仿真实验,将所提方法与灰狼优化算法（GWO）优化VMD联合小波阈值和麻雀搜索算法（SSA）优化VMD联合小波阈值两种方法对比,所提方法去噪后信号的信噪比分别提高了1.27%、2.01%,表明所提方法的去噪效果具有一定的优越性,为后续管道泄漏计算定位奠定了良好的基础。     

基于最优小波包Shannon熵的再制造电机转子缺陷诊断技术

为了实现对再制造电机转子质量的有效监控,采用超声波技术对其内部缺陷进行检测与评价。引入滤波效果良好的最优小波包滤波法和对超声波信号噪声含量变化敏感的Shannon信息熵算法,定义一种新的最优小波包Shannon熵(Best wavelet packet Shannon entropy,BWPSE)的概念,并提出基于BWPSE的超声波信号消噪方法。对采集到的再制造电机转子超声检测信号进行最优小波包滤波处理,得到各尺度的小波包分解系数,在此基础上计算各尺度小波包分解系数的Shannon熵,通过分析小波包系数Shannon熵的变化规律确定分解层数及阈值。采用该方法对再制造电机转子超声检测信号进行消噪处理,结果表明该方法对噪声消除比较彻底,对比Sqtwolog阈值小波分析及Heursure阈值小波分析等其他信号消噪方法,该方法可显著提高再制造电机转子内部缺陷定量分析的准确度。     

基于改进经验小波变换及互谱相位差谱的供水管道泄漏声振动定位方法

针对供水管道泄漏声振动信号的信噪比较低导致用于时延估计泄漏定位误差大的问题,提出基于改进经验小波变换及互谱相位差谱的供水管道泄漏定位方法。首先采用小波包分解得到不同尺度的信号能量谱,根据小波包能量谱局部极小值的分布自适应确定频带分割区间,解决了传统经验小波变换中频谱划分问题;然后基于频带分割区间构建正交小波滤波器组对信号进行经验小波变换分解得到多个分量,根据相关系数选取有效分量,同时利用有效分量的互谱相位差谱呈水平变化的频带对有效分量信号进行带通滤波,滤除干扰噪声;最后对滤波后的信号进行互相关时延估计来确定泄漏位置。仿真及实验结果表明,该方法能够有效的实现供水管道泄漏定位,并与互相关、VMD与互谱分析相结合的泄漏定位方法相比,平均相对定位误差分别减少6.7倍和1.5倍。     

分布式光纤管道泄漏检测及预警技术研究

介绍了一种基于Mach-Zehnder光纤干涉仪原理的分布式光纤管道泄漏检预警技术,利用与管道同沟敷设的光缆中的3条单模光纤构成分布式光纤微振动传感器检测管道沿线的振动信号,可以有效地检测管道沿线所发生的泄漏和异常事件.阐述了检测系统的组成和工作原理,分析了检测系统定位结构和方法.采用小波软阈值消噪对现场检测信号进行处理,可以有效地消除噪声和提高检测灵敏度.理论分析和测试结果表明,该测试技术具有较高的测试灵敏度和定位精度.     

基于小波包和HHT变换的声发射信号分析方法

针对声发射管道泄漏检测过程中的噪声干扰问题,对基于小波包和经验模态分解(EMD)的声发射信号处理方法进行了研究.采用小波包分解算法和经验模态分解都可以对管道泄漏声发射信号进行分解,但分解结果却存在一定区别.EMD是近年来非平稳信号分析领域的一个突破,对管道泄漏声发射信号进行EMD分解后,选择包含声发射特征的若干固有模式函数(IMF分量)进行重构,可以提取到管道泄漏声发射信号的本质特征,消除噪声信号的干扰.相对小波包分解方法而言,对根据IMF分量重构的声发射信号进行相关分析计算,得到的管道泄漏点的位置更为精确.     

进化小波消噪方法及其在滚动轴承故障诊断中的应用

阈值是小波阈值消噪方法中决定消噪结果的关键因素,传统阈值估计方法存在抑制噪声污染与保留信号细节间的矛盾,难以实现对滚动轴承故障信号的有效消噪.为准确估计阈值以改善小波阈值消噪方法的消噪性能,提出一种基于小波变换的进化阈值消噪方法.该方法以小波变换作为含噪信号分解与重构工具,构造含噪信号在各小波分解尺度上硬阈值收缩均方误差的近似函数,利用粒子群优化进化搜索与其最小值对应的最优阈值,以近似实现均方误差最小意义下的最优消噪.模拟信号消噪分析与滚动轴承故障信号消噪实例表明,该方法可有效消除噪声对信号的干扰,并准确提取淹没在噪声背景中的故障特征,消噪性能在信噪比与均方误差意义下优于传统小波阈值消噪方法.     

基于瞬态压力波法的输油管道泄漏检测技术研究

近年来,油管道泄漏事故频繁发生,为保障管道安全运行和将泄漏事故造成的危害减少到最小,需要研究泄漏检测技术以获得更高的泄漏检测灵敏度和定位精度。针对输油管道的特点,分析了瞬态压力波泄漏检测技术的定位原理以及影响准确定位的关键因素,对热油输送管道沿程瞬态压力波传播速度的变化进行了研究。另外利用信号奇异点与小波变换模极大值在多尺度上变化对应的性质,将小波变换应用到扑捉泄漏压力信号序列的特征拐点中,使压力波时间差更加精确,从而有效提高了输油管道泄漏检测的定位精度。     

基于小波分析的齿轮箱振动信号消噪处理

基于小波变换具有良好时频分析的特点,介绍了小波变换用于去除噪声的原理和方法,建立齿轮箱齿轮人工点蚀故障模拟实验台,通过Matlab对故障振动信号进行小波分析,并且和没有经过信号消噪就进行小波分解进行对比,从而得出对含噪声的非平稳信号进行消噪,再进行小波分解可以较好的提取齿轮点蚀故障特征频率的结论。     

基于稳态模型的气体管道泄漏定位方法的研究

基于管道中气体流动的稳态模型,根据对温度的处理方式,分别提出气体管道泄漏定位的等温定位法和变步长龙格-库塔法。通过实际气体管道泄漏的实验验证对模型法的应用场合进行了分析,并对两种方法的定位效果进行了对比。通过非等温天然气管道泄漏的仿真试验,对温差大小、气体特性估算的准确度和算法中模型的简化程度对定位效果的影响进行了研究。仿真研究表明,管道两端气体特性估算的准确程度决定了算法的定位精度,而算法中的管道模型可以进行适度的简化。此外,变步长的龙格-库塔法在非等温管道上的泄漏定位效果要优于等温定位法。     

小波在输油管道泄漏检测定位系统中的应用研究

提出一种小波变换与负压波相结合,使用阈值降噪理论对输油管道泄漏点进行检测定位的方法。实验证明,基于此法的输油管道泄漏检测定位系统切实可行,极大地提高了信号的可检测性,并能提高定位精度、减少误报率。     

基于EWT及互时频的天然气管道泄漏定位

针对天然气管道泄漏定位的问题,提出一种基于经验小波变换(EWT)及互时频的泄漏定位方法。该方法首先采用EWT对泄漏信号进行分解,获得多个分量,进而提出根据互信息的敏感分量自适应选择算法,获取敏感分量;然后采用互时频法对敏感分量进行时频分析,根据互时频图的最大特征峰值计算时延和相关频率;最后根据频散曲线获得相关声波速度,实现对天然气管道泄漏点定位。实验结果表明,该方法能够实现管道泄漏点定位,与基于EMD的互时频法相比,定位精度明显提高。     

基于Markov特征的油气管道泄漏检测与定位方法

针对传统的基于压力信号的管道泄漏检测方法误报率和漏报率偏高,同时定位误差较大的缺点,设计了一种基于Markov特征的管道泄漏检测与定位方法。首先,将管道压力数据构造为Markov链的形式,并提取其动态特征;然后,将所提取的特征应用于Neyman-Pearson异常检测方法之中,检测全部压力数据样本的状态,并对检测到的异常样本进行同源信号匹配,修正检测结果;最后,将相似性定位方法与连续小波定位方法结合,确定管道首末两端响应压力变化的时间差,并根据管道长度和压力波传输速度等信息,对泄漏源定位。所提方法能应用于小泄漏和缓慢泄漏的检测与定位,易于实现,误报率与漏报率显著降低,定位精度提高。通过对历史数据的分析,验证了所提方法的可行性和有效性。     

Gas leak locating in steel pipe using wavelet transform and cross-correlation method

Pipelines leakage in power plant, petrochemical complexes, and refineries can lead to explosion, pollution, and severe physical damages, so precise and on time leak locating is very important. There are many techniques for detecting and locating the leakage. In this research, we represent the leak locating principle in pressurized gas pipelines by using acoustic emission theory. An algorithm for finding the location of continuous acoustic waves resulted from leakage is provided by MATLAB software. The used leak locating technique is a combination of wavelet transform, filtering, and cross-correlation methods. The resulted acoustic emission signals were analyzed into high and low frequencies by wavelet transform and available noises on them were omitted completely by filtering. Then de-noised acoustic emission signals were reconstructed. Time differences of de-noised waves were calculated precisely by using cross-correlation function. For studying the accuracy of used method, acoustic emission testing was done by continuous leakage source. The resulted signals of leakage were recorded by two acoustic sensors in two sides of leakage source, and time difference and leak location were calculated by using the algorithm. Several tests were done by changing sensor distance from leakage source and error percent of less than 3 % was resulted in leak locating that indicated high precision of used algorithm.     

Adaptive noise cancellation based on EMD in water-supply pipeline leak detection

In water-supply pipeline leak detection and location, both the leak signals and blurred noises are closely related to the pipeline states and surroundings and most of the conventional noise-cancellation methods have to depend on the empirical parameters of either signals or noises. EMD (Empirical Mode Decomposition) is an adaptive signal decomposition method and is exclusive of base functions. A signal is decomposed into several IMFs (Intrinsic Mode Functions) in EMD, then the noise in a signal can be cancelled through removing uncorrelated IMFs. The existing EMD noise cancellation methods need to know the characteristics of either the wanted signal or the noise for rebuilding the noise-removed signal. However the characteristics of leak signals and noises are not fixed in various pipeline conditions, so the existing EMD noise cancellation methods can’t be directly applied in water-supply pipeline leak detection. This paper proposes an adaptive noise cancellation method based on EMD, in which the IMFs that don’t or less contain the components related to the leak can be removed through the cross-correlation between the IMFs and another signal collected at the either side of a suspect leak. In simulation analysis, the adaptive noise cancellation method can increase the SNRs (Signal to Noise Ratios) of leak signals as high as 16 dB. In processing practical pipeline vibro-acoustic signals, with the proposed method the peak of adaptive time delay estimate of leak signals, which determines the location of a leakage, becomes more distinguished, and thus the error of leakage location is improved.     

Leak detection monitoring system of long distance oil pipeline based on dynamic pressure transmitter

A dynamic pressure transmitter (DPT) is designed in this paper for long distance oil and gas pipeline, of which the design principle and main performance indices are introduced. A novel leak detection monitoring system (LDMS) of long distance oil pipeline based on DPT is designed, and the detection principle and system composition are specified. Dynamic pressure signals along the pipeline can be obtained by the DPT, and then the pipeline leak can be detected by extracting the wavelet packet entropy (WPE) of the signals. The WPE signal feature extraction method for judgment of pipeline leak is explored and developed. In addition, the influence of the wavelet basis and calculating window width on the identification performance of the WPE are further discussed. The application examples show that the DPT has higher detection sensitivity and leak resolution than the traditional pressure transmitter. The system can identify the pipeline leak correctly and reduce false alarm rate effectively. It can also improve detection sensitivity and localization accuracy. The system can correctly detect leaks with a rate of 96.7% and the greatest localization error is 101 m.     

基于LMD包络谱熵及SVM的天然气管道微小泄漏孔径识别

针对管道泄漏信号的非平稳特征以及管道泄漏孔径大小难以识别的问题,提出一种基于局域均值分解包络谱熵及支持向量机的识别方法。该方法对管道泄漏信号进行局域均值分解,得到若干个瞬时频率具有物理意义的乘积函数(production Function, PF)分量;计算各PF分量的峭度值并据此选出包含主要泄漏信息的分量作为主PF分量,对这些分量进一步采用小波包分解能量法进行分析并重构;再对重构后的主PF分量进行希尔伯特变换求取包络谱,结合信息熵的概念提出包络谱熵并计算熵值;将归一化包络谱熵作为泄漏信号特征输入支持向量机分类器中,用以区分不同的泄漏孔径,完成对泄漏孔径的识别。通过试验采集大量的管道泄漏信号进行处理及分析,试验结果表明该方法能有效识别不同泄漏孔径类别。     

气体管道泄漏模态声发射时频定位方法

针对声发射信号频散特性导致基于时延估计的气体管道泄漏定位误差大的问题,提出一种基于模态声发射时频分析的泄漏定位方法。该方法采用平滑伪Wigner-Ville时频分布对两泄漏信号的互相关函数进行时频分析,利用互相关函数的时频谱可同时提取泄漏信号的时间延迟和与之对应的频率;然后根据泄漏声发射信号的主导模态的频散曲线即可确定该频率对应的声速,利用实时确定的声速和时间延迟并根据两传感器之间的距离即可确定泄漏点的位置。实验结果表明,采用时频分析的气体管道泄漏定位误差与互相关相比减少了6倍。所提出的模态声发射时频定位方法能有效抑制泄漏信号的频散,提高泄漏信号的相关性,从而更适合用于声发射管道泄漏定位。     

A new denoising algorithm for MFL data obtained from seamless pipeline inspection

The magnetic flux leakage (MFL) nondestructive evaluation technique is used extensively for in-service inspection of gas and oil pipelines. Unfortunately, the MFL data obtained from seamless pipeline inspection is usually contaminated by various sources of noise, which considerably reduces the detectability of defect signals in MFL data. In this paper, a new denoising algorithm is presented for removing seamless pipe noise (SPN) and system noise contained in MFL data. The algorithm first utilizes the new wavelet domain adaptive filtering method proposed by combining wavelet transform with the adaptive filtering technique to remove SPN contained in MFL data and then exploits the coefficient denoising approach with wavelet transform to cancel the system noise in the output of the wavelet domain adaptive SPN cancellation system. Theoretical analysis shows that the proposed denoising algorithm has a better overall performance than the existing denoising algorithm. Results of application of the proposed algorithm to MFL data from field tests are presented to demonstrate the performance of the proposed algorithm compared with the existing denoising algorithm. The text was submitted by the author in English.