激光超声信号去噪的经验模态分解实现及改进

考虑激光超声检测过程中噪声对缺陷和材料特征分析和检测的影响,本文以激光超声信号去噪为目的,研究了基于经验模态分解（EMD）的激光超声信号时间尺度滤波过程。针对分解过程中固有模态函数（IMF）上有用信号与噪声的混叠现象对重构信噪比的影响,结合信号多模态和宽频带的特点,提出了基于峰度检验策略的时域加窗方法。该方法通过局部峰度检验判断重构起点附近IMF中有用信号的位置及信噪分界点,利用Turkey-Hanning窗保存有用信号,抑制噪声,实现信号与噪声的解混叠,改善重构信号质量。仿真和实验结果表明,该方法具有良好的自适应性,有效识别并分离了信号和噪声成分,信噪改善比达14 dB以上,相对原始方法提升了3 dB,相对性能增强了20%,并且改进效果随信号受污染程度的加重而愈发突出,有望在高噪声水平下发挥优势。     

声发射无损检测技术在管道故障检测中的应用

管道运输工作现场环境复杂,复杂多变的噪声源影响所监测到测量信号,针对复杂信号的提取。本文在退役管道上通过模拟实验获取不同类型的三类信号：断铅、金属棒敲击和砂纸三种声发射信号源,获取故障信号数据,并将数据进行处理,成为可显示的电子信号图像,从而可以进行采集,处理和分析,同时,因为信号具有复杂性,提出对采集到的信号进行小波去噪,去除信号中无意义的部分,大量采集信号中所需要的的信号,为故障信号数据类别判定提供有效保障。     

基于多重小波变换的信号去噪及其在软测量中的应用

针对化工过程中数据的随机性、非平稳性及含有大量噪声的特点,提出了小波去噪的思想和步骤,并在此基础上采用多重小波变换阈值去噪的方法,以去除大部分高频随机噪声,提高数据的置信度。然后将该方法应用于醋酸正丙酯反应过程质量指标软测量模型中,仿真结果表明,该方法能有效恢复数据的真实性,能提高软测量建模的拟合精度与泛化性能。     

管道腐蚀缺陷超声信号的PSO-SVM模式识别研究

针对金属管道腐蚀问题,提出了一种基于支持向量机(Support vector machine,SVM)与粒子群优化(Particle swarm optimization,PSO)相结合的管道腐蚀缺陷的分类方法。对预处理后的超声缺陷信号进行经验模态分解(Empirical mode decomposition, EMD),提取相应的时域无量纲参数作为特征向量;建立SVM缺陷分类模型,并采用PSO算法优化SVM参数,提高模型的缺陷分类准确率。实验证明,该方法建立的模型针对不同深度的超声缺陷信号的识别率达到87.5%,优于相同试验样本下BP神经网络和RBF神经网络的分类准确率。     

壳段厚度激光检测信号的变分模态分解去噪

针对双激光位移传感器测量大型壳段厚度过程中噪声对检测精度的影响,提出利用变分模态分解来实现对厚度信号的自适应去噪,利用相邻固有模态函数之间的离散Hellinger距离来获取最佳的模态数。该方法将变分模态分解算法引入到激光信号的自适应滤波过程中,分析并改进了变分模态分解算法的过分解、欠分解以及能量泄露的问题。然后,对改进的变分模态分解与希伯特振动分解和自适应噪声总体集合经验模态分解进行性能对比,提出了固有模态函数的相对瞬时能量概率的概念。最后,结合离散Hellinger概率分布距离理论判断固有模态之间的信噪分界点,实现了对信号的重构及滤波处理。仿真和实验结果表明,该方法对壳段厚度信号处理的信噪比为39.27dB,比自适应噪声总体集合经验模态分解方法提高了10dB,具有良好的自适应性,无需先验条件便能快速有效地识别并分离激光信号中的噪声成分。     

提升小波变换在振动信号去噪中的应用

振动信号存在不同形式的波形特征,传统小波去噪中,小波分解的结果与所采用的小波基函数有关,选用不适当的小波基函数会冲淡振动信号的局部特征信息,从而造成原始信号的部分有用信息丢失。为了克服上述缺陷,介绍了提升算法和基于该算法的小波变换快速算法,探讨了如何利用提升小波变换对信号进行去噪一通过对实际信号去噪处理．得出了提升小波算法能够较好地应用于信号去噪的结论。     

超声检测信号中的小波包分析

超声检测中缺陷回波信号的质量是实现缺陷定位定量定性分析与评价的基础,去噪是超声检测信号处理中很重要的环节。小波包分析方法可以有效地抑制噪声干扰,提离有用信号。在Windows环境下,采用MATLAB 6．5 编程实现了缺陷回波信号的小波包去噪处理。研究结果表明:利用小波包阈值去噪,大大提高了缺陷回波信号的信噪比,改善了信号特征信息的质量,并且具有较高的缺陷定位精度和纵向分辨率。     

超声信号的特征提取与选择在缺陷分类识别中的应用

超声缺陷回波信号的特征提取与选择是超声波检测的基础和关键。结合待检测超声信号的特点,在时域、频域上研究了超声缺陷信号的特征提取,提出SBS和SFS结合的特征选择方法。在满足识别准确率的同时,有效地降低了分类器输入向量的维数,提高了运算效率。     

基于多重小波变换的信号去噪及其在软测量中的应用

化工生产过程中采集到的数据信号通常具有随机性和非平稳性，附加了各种噪声，以至于影响数据建模的拟合效果和泛化性能。本文基于小波分析的特点，提出了一种对信号数据进行多重小波变换阈值去噪的方法。该方法可去除大部分高频随机噪声，提取真实信号，进而提高数据的置信度。将该方法与小波神经网络相结合并应用于丙烯腈聚合反应过程质量指标软测量模型中。仿真结果表明，该方法能有效恢复数据的真实性，提高数据建模的拟合精度与泛化性能。     

单频激光干涉仪正交信号的高精度处理

为了提高单频激光干涉仪正交信号相位细分辨向的可靠性与重复性,本文在正交信号Heydemann误差模型和数字信号处理技术的基础上,提出一种结合误差修正和相位细分辨向技术的正交信号高精度误差补偿算法。该算法采用基于最小二乘法的矩阵运算计算正交误差补偿参数初值,通过迭代运算进一步提高补偿精度,并对修正后的信号构建了基于相位的细分辨向算法。最后通过MATLAB软件对该算法进行了验证。实验结果表明,上述算法可实现对正交信号误差的精确补偿,使测量精度可达亚纳米甚至皮米数量级,从而有效提高测量信号的解调精度。     

Wind speed and direction measurement based on arc ultrasonic sensor array signal processing algorithm

This article investigates a kind of method to measure the wind speed and the wind direction, which is based on arc ultrasonic sensor array and combined with array signal processing algorithm. In the proposed method, a new arc ultrasonic array structure is introduced and the array manifold is derived firstly. On this basis, the measurement of the wind speed and the wind direction is analyzed and discussed by means of the basic idea of the classic MUSIC (Multiple Signal Classification) algorithm, which achieves the measurements of the 360° wind direction with resolution of 1° and 0–60 m/s wind speed with resolution of 0.1 m/s. The implementation of the proposed method is elaborated through the theoretical derivation and corresponding discussion. Besides, the simulation experiments are presented to show the feasibility of the proposed method. The theoretical analysis and simulation results indicate that the proposed method has superiority on anti-noise performance and improves the wind measurement accuracy.     

A novel ultrasonic array signal processing scheme for wind measurement

This article presents a novel ultrasonic array signal processing scheme for wind speed and wind direction measurement. Two ultrasonic array structures are illustrated and the corresponding array signal processing method is proposed. The statistical performance of proposed scheme, such as estimation variance and Cramer Rao Bound (CRB), is derived. The ambiguity problem for traditional uniform line array (ULA) structure is theoretically discussed. The hardware experiment results verify the feasibility of the proposed wind measurement method.     

基于BP神经网络的管道泄漏声信号识别方法研究

针对城市供水管网泄漏检测需求,进行了泄漏声信号识别方法研究。分析了泄漏信号的时域、频域及波形特点,提取出可用于泄漏信号表征的20种特征参数;基于提取的泄漏声信号特征参数,构建了泄漏声信号BP神经网络识别系统;研究了神经网络结构(隐含节点数、传递函数、学习率)及输入参数的数量和种类对泄漏信号识别效果的影响,并优化出最佳的神经网络结构及输入参数。在以上研究基础上,利用优化后的神经网络对实验室及现场管道泄漏信号进行了交叉训练和识别,结果表明,提出的基于泄漏特征参数的神经网络系统具有较高的可靠性和普适性,可以很好地实现不同场景下泄漏信号的交叉识别,整体识别率达92.5%,为解决不同工况下泄漏信号识别做了有益的探索。     

基于多信号流图与差分进化算法的测点布局优化

为提高系统可测试性设计过程中的测点布局优化效率,提出了一种基于多信号流图与差分进化算法的测点布局优化方法。该方法首先建立基于多信号流图的系统模型,随后根据系统模型获得测试与故障模式的依赖矩阵,最后根据测点布局对探测率、隔离率和测点数量等方面的灵活需求,通过依赖矩阵和差分进化算法寻找最优测点组合。仿真实验和真实应用案例均证明了该方法的有效性。同时,与基于遗传算法的同类方法的对比实验,还证明了本文方法能更快且更稳定地找到最优测点组合,因此更适用于大型复杂系统的设计。     

随钻姿态测量重力加速度自适应提取算法

为解决旋转导向钻井过程中由于钻具的高速旋转、剧烈振动和冲击给随钻姿态测量带来巨大干扰的问题,基于互补滤波框架提出了一种钻具重力加速度提取算法。首先通过余弦矩阵微分方程得到基于陀螺仪的重力加速度的递推方程,然后利用陀螺仪和加速度计在频率上互补的特性,在互补滤波框架下融合得到钻具重力加速度;同时,为了提高系统鲁棒性,利用非重力加速度和钻具旋转速度作为模糊算法输入,通过模糊算法动态调整互补滤波参数;最后设计了振动和旋转实验验证了算法的有效性。实验结果表明,使用本文算法提取重力加速度后姿态估计精度得到明显提升,本文算法相比于基于Kalman滤波框架的重力加速度提取算法在PC机测试环境下单次运算时间减少31.9%。     

Collection deficiencies of scanning electron microscopy signal contrasts measured and corrected by differential hysteresis image processing

Limitations of scanning electron microscopy (SEM) image resolution and quality were measured in digital image data and their effect on image contrasts was analyzed and corrected by differential hysteresis (DH) processing. DH processing is a mathematical procedure that utilizes hysteresis properties of intensity variations in the image for a segmentation of differential contrast patterns. These patterns display contrast properties of the data as coherent full-frame images. The contrast segmentation is revertible so that the original image can be restored from the sum of the sequentially extracted DH contrast patterns. DH imaging enhances weak contrast components so that they are more easily recognizable and displays SEM image data free of signal collection efficiency contrasts. Example image data include environmental SEM (ESEM) and SEM images of low and mediumhigh magnifications where collection deficiencies included charging of the specimen surface, obstructions from specimen topography, and uneven signal collection properties of the detector. ESEM low-vacuum image data, which appear to be of high quality, contained local areas of reduced contrasts due to residual surface charging. In such areas, signal contrasts were reduced up to 80%, which suppressed most of the weak short-range contrasts. In low-magnification SEM images, up to 93% of the local high precision contrast was lost from the various adverse effects which diminished the pixel-related contrast resolution of the microscope and resulted in images with low detail. Also, at medium magnification, surface charging effects dramatically reduced the image quality because contrasts resulting from local electron beam/specimen interactions were reduced by as much as 71%. DH imaging restored the local contrast losses by elimination of the collected distorted fraction of signal contrasts and reconstitution of the collected maintained fraction. Restored DH images are of superior quality and enhance the imaging capability of the conventional SEM. DH contrast segmentation provides an improved basis for the measurement of various signal contrast components and detector performances. The DH analysis will ultimately facilitate a precise deduction of specimen properties from extracted contrast patterns.