Signal detection and noise suppression using a wavelet transform signal processor: application to ultrasonic flaw detection

The utilization of signal processing techniques in nondestructive testing, especially in ultrasonics, is widespread. Signal averaging, matched filtering, frequency spectrum analysis, neural nets, and autoregressive analysis have all been used to analyze ultrasonic signals. The Wavelet Transform (WT) is the most recent technique for processing signals with time-varying spectra. Interest in wavelets and their potential applications has resulted in an explosion of papers; some have called the wavelets the most significant mathematical event of the past decade. In this work, the Wavelet Transform is utilized to improve ultrasonic flaw detection in noisy signals as an alternative to the Split-Spectrum Processing (SSP) technique. In SSP, the frequency spectrum of the signal is split using overlapping Gaussian passband filters with different central frequencies and fixed absolute bandwidth. A similar approach is utilized in the WT, but in this case the relative bandwidth is constant, resulting in a filter bank with a self-adjusting window structure that can display the temporal variation of the signal's spectral components with varying resolutions. This property of the WT is extremely useful for detecting flaw echoes embedded in background noise. The detection of ultrasonic pulses using the wavelet transform is described and numerical results show good detection even for signal-to-noise ratios (SNR) of -15 dB. The improvement in detection was experimentally verified using steel samples with simulated flaws.     

Improved ultrasonic flaw detection technique for austenitic stainless steel welds

Conventional ultrasonic techniques are seldom employed for evaluating austenitic stainless steel welds. This is because the coarse dendritic structure of the austenitic stainless steel welds results in high attenuation of the ultrasonic waves and also produces spurious echoes called ‘grass’. As a consequence, the effective detection of real defects is impaired by the poor signal to noise ratio. A modified version of the conventional technique, which overcomes these problems to a large extent, is presented. Results obtained on a few sample welds are also presented to demonstrate the efficacy of the technique.     

车轮电磁超声探伤技术的研究

我国铁路相关部门对车轮的检测主要采取定期检修的方式，即在规定的时间里集中大量的人力，对车轮有计划地进行检修，这样就需要很大的人力和物力。为了改善这种情况，我们研制了车轮动态自动探伤系统，本文主要介绍电磁超声表面波的激发和接收机制，利用电磁超声表面波对车轮探伤的原理，并分析了一些重要的探伤参数和缺陷波的处理方法。     

Analysis of order-statistic CFAR threshold estimators for improved ultrasonic flaw detection

In the pulse-echo method using broadband transducers, flaw detection can be improved by using optimal bandpass filtering to resolve flaw echoes surrounded by grain scatterers. Optimal bandpass filtering is achieved by examining spectral information of the flaw and grain echoes where frequency differences have been experimentally shown to be predictable in the Rayleigh scattering region. Using optimal frequency band information, flaw echoes can then be discriminated by applying adaptive thresholding techniques based on surrounding range cells. The authors present order-statistic (OS) processors, ranked and trimmed mean (TM), to robustly estimate the threshold while censoring outliers. The design of these OS processors is accomplished analytically based on constant false-alarm rate (CFAR) detection. It is shown that OS-CFAR and TM-CFAR processors can detect flaw echoes robustly with the CFAR of 10 (-4) where the range cell used for the threshold estimate contains outliers.     

Near surface flaw detection by ultrasonic critical angle imaging

Acoustic critical angle measurements usually involve broad beam or global isonification of a water-solid interface. In order to obtain local interface data, it is convenient to use a focused acoustic source with lens axis at incident angle and point-receiver at specular reflection angle . By scanning this combination at various critical angles (=_crit), interface waves are generated locally near the lens focus. These waves penetrate the solid to a depth of approximately one shear wavelength _T and are modulated by discontinuities such as flaws, inclusions, and debonds that may be present there. A radiated longitudinal wave in water accompanies these modulated interface waves and carries information on subsurface flaws to the point-detector along the direction of a specularly reflected wave. By scanning the focus-receiver combination along the interface, the receiver output may be used to form images of local subsurface detail. We show that in anisotropic materials, such as crystals, there are generally two different _crit where a radiating interface wave provides subsurface information. We also discuss special problems associated with the use of a focused acoustic source including nonlinear effects (generation of harmonics) in the water.     

Signal space detection in colored noise

An appropriate adaption of the decision boundaries of the well-known signal space detector (SSD) yields a detector that whitens the input noise in the detector forward path. This new detector, called a whitening signal space detector (WSSD), offers higher reliability without increasing the dimensionality of the signal space. A WSSD can be designed by applying a transformation into the ordinary SSD case. We use the new concept to design a WSSD based on three-dimensional 110 equalization and demonstrate its feasibility and performance. The detector can be implemented with a small increase of hardware and offers a significant improvement in terms of bit error rate, especially at low to moderate channel densities     

Metrological aspects of calibration of standard test blocks for ultrasonic flaw detection

Adjustment of ultrasonic detection sensitivity with the aid of standard tes blocks with noncircular reflecting surfaces is discussed. The shapes of reflecting surfaces and their angular position in the block with respect to the plane of entry of ultrasonic oscillations make it possible to use them for adjusting straight and oblique transducers. Composite test blocks for inspecting plane and rod-shaped materials are described. Translated from Izmeritel'naya Tekhnika, No. 5, pp. 35–38, May, 1997.     

超声水浸探伤检测闸门自动跟踪电路研究

针对传统超声探伤应用中缺陷信号选通闸门通过硬件或程序预先设置,当缺陷信号相对闸门位置发生变化而引起系统漏检或误判的问题,设计一种高可靠性的闸门自动跟踪电路。该电路具有自动跟踪水钢界面波位置,实时调整选通闸门开启位置选择对应缺陷波的功能。通过和常规闸门电路进行对比测试,该跟踪电路性能稳定、可靠,能够自动跟踪超声检测的界面波并自动调整闸门位置对缺陷波进行探伤。经过实际检测测试:该电路能有效将检测系统中由于工件位置相对浮动而产生的漏检或误判概率降低一个数量级,为精确判伤提供可靠保障。     

小波域的频率扩展信道的信号检测

由于海洋动力学和目标起伏引起发射信号包络的时变衰减，相应于频域就是频率扩展，这种信道称为频率扩展信道，又名快速衰落信道(FFD)。在FFD信道中，理想信道的最佳检测器——匹配滤波器的性能将有所下降，Baggenstoss已经证明FFD信道的最佳检测器是分段副本相关器(SRC)。近年来出现的小波变换具有和宽带相关处理相似的计算结构，所以文章拟从连续小波变换的理论和性质出发，推导FFD信道小波变换域最佳检测器的形式，并针对一个仿真算例，进一步用Monte-Carlo实验计算出该检测器的性能曲线，结果表明该检测器的性能优于小波域的副本相关器。     

基于复小波变换和支持向量机的缺陷类型识别

 针对传统缺陷检测存在的工序繁琐、不易在线实施、准确率低、容易受人为因素影响,以及用人工神经网络对小样本事件进行缺陷类型识别时存在泛化能力差和过学习等问题,提出一种基于复小波变换和支持向量机（SVM）模式识别理论的缺陷类型识别新方法．在利用小波对超声缺陷回波信号进行消噪的基础上,采用复小波变换获得缺陷回波信号的包络并提取其特征参数,构成输入特征向量后运用支持向量机进行分类．实验结果表明,该方法具有识别准确率高、泛化能力强、容易实现在线处理等优点．     

Narrowband impedance matching layer for high efficiency thickness mode ultrasonic transducers

A new matching layer design concept has been proposed for narrowband continuous wave (CW) devices. Analysis has shown that the mechanical impedance of a resonant-type transducer in thickness mode CW operation does not equal its acoustic impedance ρVs but roughly equals ρVs/Q, where ρ is density, Vs is acoustic velocity, and Q is the mechanical quality factor. The value of ρVs/Q is much lower than the acoustic impedance of water for any transducer material, including lead zirconium titanate (PZT), single crystals, or polyvinylidene fluoride (PVDF). With this new approach, the impedance of the matching layer must also be between water and ρVs/Q, but there are few such practical low impedance materials. To realize equivalent low impedance structure, a novel double layer design is presented: a relatively low impedance material (such as polyethylene or polyurethane) on the inside and a relatively high impedance material (such as polyester or metal) on the outside. A high power CW transducer structure was designed and fabricated with PVDF-TrFE (polyvinylidene fluoride trifluoroethylene) to operate at 1.4 MHz     

Analysis of order statistic filters applied to ultrasonic flaw detection using split-spectrum processing

Split-spectrum processing of broadband ultrasonic signals coupled with order statistic filtering has proven to be effective in improving the flaw-to-clutter ratio of backscattered signals. It is shown that an optimal rank can be obtained with a prior knowledge of flaw-to-clutter ratio and the underlying distributions. The order statistic filter performs well where the flaw and clutter echoes have good statistical separation in a given quantile region representing a particular rank (e.g. minimum, median, maximum). Order statistic filters are analyzed for the situation in which the observations do not contain equivalent statistical information. Experimental and simulated results are presented to show how effectively the order statistic filter can utilize information contained in different frequency bands to improve flaw detection.     

A unified constrained inversion model for ultrasonic flaw sizing

Equivalent flaw sizing using ultrasonic waves is an approach whereby shape and orientation information of a defect are obtained in terms of a best-fit simple geometry that is able to represent the major aspects of the flaw. Separate examples of this approach have previously been developed for volumetric flaws and cracks using the Born and Kirchhoff approximations, respectively. Here, these separate algorithms are unified into a single algorithm capable of sizing both volumetric flaws and cracks. Some examples of the performance of this unified algorithm on both synthetic and experimental data are also given.     

Two-dimensional inverse born approximation in ultrasonic flaw characterization

A method is developed to characterize flaws of arbitrary shape by using ultrasound pulse echoes at multiple coplanar incident directions. The three-dimensional image reconstruction problem is reduced to a series of two-dimensional image reconstructions, thereby avoiding the difficulties associated with three-dimensional image reconstructions, such as taking and processing a large amount of data, and the complications associated with three-dimensional image reconstructions, such as three-dimensional interpolation, long computing time, etc. The reconstructed two-dimensional images represent the two-dimensional projections or shadows of the three-dimensional flaw characteristic function. Each projection image is reconstructed independently using well-developed computerized tomography reconstruction techniques. If the shape of the flaw is not too irregular, or if the fine details of the shape are not of interest, only a few of these projection images suffice to characterize the flaw. The magnitude scaling problem and the alignment problem of the echoes at different incident directions can be handled easily in the algorithm. Simulation studies yielded encouraging results.     

Quantitative spectral analysis for flaw detection in concrete

The resonance method of transient stress wave propagation is employed for the detection of flaws in concrete. Quantitative analysis of the spectra identifies the resonance mode due to reflection from the flaw, and enables the exclusion of the flexural modes of the concrete plate above the flaw. Two-dimensional scanning of a test point provides the information about the depth of internal flaws. The test results of a large concrete block containing several types of artificial flaws are presented.     

Signal detection in power-law noise: effect of spectrum exponents

Many natural backgrounds have approximately isotropic power spectra of the power-law form, P(f)=K/f(beta), where f is radial frequency. For natural scenes and mammograms, the values of the exponent, beta, range from 1.5 to 3.5. The ideal observer model predicts that for signals with certain properties and backgrounds that can be treated as random noise, a plot of log (contrast threshold) versus log (signal size) will be linear with slope, m, given by: m=(beta-2)/2. This plot is referred to as a contrast-detail (CD) diagram. It is interesting that this predicts a detection threshold that is independent of signal size for beta equal to 2. We present two-alternative forced-choice (2AFC) detection results for human and channelized model observers of a simple signal in filtered noise with exponents from 1.5 to 3.5. The CD diagram results are in good agreement with the prediction of this equation.