小波神经网络在超声检测信号中降噪的研究

摘 要: 为了最大限度的消除粗晶材料超声检测时,晶粒散射波对有用信号的严重干扰,提高接收信号的信噪比,将小波神经网络引入粗晶材料超声检测信号处理领域中。在训练小波神经网络时,采用了改进的梯度下降算法。该网络有一个动态的权值,它随误差变化而调整。结果表明,小波神经网络应用在粗晶材料超声检测信号的降噪时,能够达到较理想的降噪效果。     

粗晶材料超声无损检测信号处理新技术

在超声检测中,粗晶材料由于较大的噪声背景,严重影响着有用信号的检测,为了提高信噪比,本文介绍了几种新的粗晶材料检测信号的处理技术,包括相关技术、专家系统、小波包去噪技术以及神经网络技术,给出了几种技术的原理及应用,这些技术将推动超声检测信号处理技术向着实时、高效的方向发展。     

几种超声NDE信号降噪方法的性能比较

为提高超声无损检测的准确性,需要对超声NDE信号中因随机分布于媒质中的大量散射微粒所引起的结构噪声进行降噪。由于信号和噪声的频谱范围基本重叠,传统的线性滤波方法不能提供理想的降噪结果。介绍了几种对超声NDE信号进行降噪的新方法:Wigner-Ville分布法、小波变换法和基于时间延迟的神经网络法,并从信噪比(SNR)、检测概率(PD)和估测深度(ED)等三个重要参数对它们的降噪性能进行计算机仿真实验的比较。结果表明:小波变换法和神经网络法的降噪效果较Wigner-Ville分布法要好。对实际信号的测试还表明,小波变换由于不像神经网络那样需要训练,是一种更为理想的超声NDE信号降噪方法。     

基于小波变换的低碳钢连铸坯超声探伤

利用离散小波变换对粗晶钢材超声检测信号进行处理,比较了使用不同的小波基,并对不同尺度中突出体现的信息分别进行处理后重构信号的去噪效果。结果表明,对不同尺度系数作不同的阈值处理有利于缺陷的识别,这种方法对粗晶钢材料的检测有积极的意义。     

截止频率法提取粗晶材料超声检测的目标反射信号

对于粗晶材料的超声检测，当组织噪声很强而目标反射信号又很弱时，分离谱谱技术及其常规的恢复算法不能直到有效的降噪作用。本文在分离频谱法的基础上，同种改进的信号处理技术－－截止频率法。实验结果表明，该方法能从很强的组织噪声中将微弱的目标反射信号提取出来，从而有效地消除了材料组织噪声的影响。     

基于水平分置线性双阵列的超声全聚焦成像方法在粗晶材料检测中的应用

针对粗晶材料超声检测信噪比低的问题,提出了一种水平分置线性双阵列超声成像方法。将两个线阵超声换能器沿直线水平分置在待检区域表面两侧,用收发分离的信号采集模式,一侧激发,另一侧记录各通道数据,进行聚焦成像。相比单阵列和同位置双线阵检测,文中的方法有效地减少了背向散射信号对缺陷信号的干扰,提高了成像信噪比。在粗晶铜质试块上的成像实验结果表明,当缺陷距离阵列较近时,文中的方法优于单阵列和同位置双线阵方法,成像信噪比提高约5～10 dB;当缺陷距离阵列较远时,单阵列模式和同位置双线阵检测方法失效,但文中的方法依然可以识别缺陷。文中的研究为粗晶材料的超声检测提供了一种可行的方案。     

小波包能量谱和BP神经网络在波纹管压浆超声检测中的应用

针对小波分析在信号处理的局限性,将小波包分析和反向传播(Back Propagation,BP)神经网络相结合,提出一种基于小波包能量谱和BP神经网络的波纹管压浆超声检测方法。采用超声检测方法接收波纹管模型的回波信号,以小波包分解后各子频带的能量作为检测特征,当波纹管内部出现脱落时,检测特征会发生变化,最后将特征输入BP神经网络中进行分类识别。试验结果表明,该方法能够理想地实现波纹管内部缺陷的诊断,可为波纹管超声检测提供一定的技术支持。     

超声多普勒信号降噪新方法的研究

本提出了利用离散小波框架结合软阈值的方法对超声多普勒信号进行降噪。通过对模拟多普勒信号和临床信号的降噪实验,发现利用小波框架的降噪方法可以获得比标准正交小波变换的降噪方法更好的降噪效果。     

基于EMD和小波熵阈值算法的超声回波信号降噪

超声检测信号中通常包含大量噪声,而其中材料晶界散射的噪声是一种相关噪声。鉴于传统的方法难以将这种噪声和缺陷回波信号区分,提出一种EMD和小波熵阈值联合降噪的算法。该算法首先对目标信号进行EMD分解,提取具有噪声特性的IMF分量进行小波分解,利用含噪系统熵增的特性,在分解各尺度层的细节部分选用小波熵自适应阈值降噪,然后将剩余分量和降噪处理后的信号进行重构。仿真信号结果表明:该降噪方法(EMD-WET)输出信号的信噪比(SNR)为7.9 d B、均方根误差(RMSE)为18.1、相似系数(NCC)为0.92,优于传统的小波软、硬阈值方法。对实测信号进行处理,该方法降低信号中的大部分噪声,更好地还原回波信号的波形。     

基于最大Lyapunov指数和小波变换的管道检测信号降噪技术

针对分布式光纤管道安全预警系统检测信号的混沌特性,最大Lyapunov指数（λm ax）被定义为降噪指标用于评价各种小波及其阈值组合的降噪性能.基于小波阈值降噪法,采用不同的小波族系和阈值选取及重调方法对现场实测人工挖掘信号进行降噪处理.现场实验数据的分析结果表明检测信号存在混沌特征.在此基础上,λm ax被用于评价小波降噪性能.最后,通过对比实验结果,sym2小波、Sqtwolog阈值选取规则和M ln阈值重调方法,可以有效地消除现场实测信号中的噪声,达到最优的降噪效果.     

Estimation of Parameters of Universal Distribution of Moments of Failure of Products

A method is proposed for estimating the parameters of a mixture of exponential and Weibull distributions for which the accuracy of preliminary estimates obtained by graphical analysis is refined in accordance with the criterion of maximum likelihood. The efficiency of the proposed method is supported by the results of statistical modeling.     

Assessment of intakes of mixtures of radionuclides

Australia has several uranium mines and a large number of mineral sand mines, with associated processing facilities. Exposures resulting from these mining and processing operations usually involve intakes of mixtures of radionuclides. This work describes the development of a suite of first order, linear compartment models, based on the ICRP Publication 66 respiratory tract model, and an analytical solution to the decay equations, for assessing the consequences of such intakes. The computer programs based on these models directly compute excretion, organ retention and organ and whole-body doses for intakes of either single radionuclides or any mixture of radionuclides belonging to the same radioactive decay chain. The intake can be via inhalation, ingestion or injection, and can be acute, chronic or of limited duration. The starting concentration and degree of secular (dis)equilibrium can be specified for each radionuclide. No assumptions need to be made about the relative magnitudes of the radioactive half-lives of the different nuclides.     

Analysis of the rate of propagation of the wave of change of boiling modes

An approximate analytical expression is derived for the velocity of the front of the wave of change of boiling modes on a rod and on a plate. The impact of the Thomson effect and of the heater orientation in the gravity field is included. Adequate agreement is demonstrated between the derived relations and experimental data.     

Evaluation of optical properties of highly scattering media by moments of distributions of times of flight of photons

A novel method for the determination of the optical properties of tissue from time-domain measurements is presented. The data analysis is based on the evaluation of the first moment and the second centralized moment, i.e., the mean time of flight and the variance of the measured distribution of times of flight (DTOF) of photons injected by short (picosecond) laser pulses. Analytical expressions are derived for calculation of absorption and of reduced scattering coefficients from these moments by application of diffusion theory for infinite and semi-infinite homogeneous media. The proposed method was tested on experimental data obtained with phantoms, and results for absorption and reduced scattering coefficients obtained by the proposed method are compared with those obtained by fitting of the same data with analytical solutions of the diffusion equation. Furthermore, the accuracy of the moment analysis was investigated for a range of integration limits of the DTOF. The moment analysis may serve as a comparatively fast method for evaluating optical properties with sufficient accuracy and can be used, e.g., for on-line monitoring of optical properties of biological tissue.     

Numerical study of methods of linearization of the output characteristics of radio-frequency sensors of the level of dielectric media

Different constructions of the sensing elements of radio-frequency level sensors based on sections of long lines are numerically investigated by means of the Matlab program for the purpose of minimizing the measurement error caused by the nonlinearity of the output characteristic. It is shown that the nonlinearity is lowest for U-shaped designs. The nonlinearity factor KN may be regulated by varying the capacitance of a correcting capacitor connected to the input of the sensing element and (or) the length of one of the sections of the long line. In selecting an optimal capacitance of the correcting capacitor KN ≤ 0.15% for control media with dielectric permittivity in the range 2.2–30.     

Possibilities of Decreasing the Errors of Determination of the Height of the Pole of a Sessile Drop of Melt

We analyze the possibilities of decreasing the errors of measuring the height of the pole of a sessile drop of melt above its equator by determining the position of the equator according to the midpoints of vertical chords of the image of the drop. The optimal distance between these chords and the edge of the image of the drop is chosen within 10–20% of the radius of the equator. The proposed method guarantees the same level of errors of evaluation of the height of the pole and the radius of the equator, which enables one to substantially increase the accuracy of evaluation of the capillary constant and the surface energy of the melt according to the geometric parameters of the sessile drop.     

Analysis of anisotropy of properties on the basis of studies of texture of coarse-grained ingots of thermoelectric materials

An analysis of anisotropy of the properties of a Bi₂Se_0.3Te_2.7 solid solution was carried out using construction of demonstrative surfaces for thermoelectric effectiveness and thermal expansion coefficients. It is shown that the texture is an important factor forming anisotropy of properties and technological fitness of ingots for manufacturing modules. Anisotropy of properties based on the studies of the ingot textures obtained using the float-zone method and Bridgman method (growing thermoelectric plates in a flat cavity) was studied.     

Estimation of potentialities of methods of summation of direct measurement errors

Conventional methods for summation of direct measurement errors are considered. Conditions under which these methods provide consistent results are stated. It is proposed to utilize for characteristic of the total error, its interval estimator or the half-sum of the value obtained by adding up its arithmetic and geometric random errors.Translated from Izmeritel'naya Tekhnika, No. 4, pp. 10–12, April, 1995.