超声检测中人工神经网络对缺陷定量评价

分析了均匀各向同性介质的超声检测中,耦合因素和仪器因素造成的脉冲回波幅度的差异,提出了误差校正方法。以校正后的缺陷回波和底面反射波的峰-峰值为特征量,利用人工神经网络进行缺陷类型识别和大小评价。
为模拟自然缺陷的二基本要素——光滑曲面和带棱边的平面,用有机玻璃制作了代表性的横穿孔和平底孔缺陷样品共18个.对18个缺陷样品的缺陷回波和底面反射波的峰-峰值测量了四次,并进行的校正.用人工神经网络对这组缺陷样品进行的处理结果表明:(1)设定的缺陷类型全部准确识别。(2)估计缺陷大小与标称孔径吻合较好。最后,对测量误差和缺陷大小估计误差进行了分析。     

Ultrasonic flaw detection using threshold modified S-transform

Interference noising originating from the ultrasonic testing defect signal seriously influences the accuracy of the signal extraction and defect location. Time–frequency analysis methods are mainly used to improve the defects detection resolution. In fact, the S-transform, a hybrid of the Short time Fourier transform (STFT) and wavelet transform (WT), has a time frequency resolution which is far from ideal. In this paper, a new modified S-transform based on thresholding technique, which offers a better time frequency resolution compared to the original S-transform is proposed. The improvement is achieved by the introduction of a new scaling rule for the Gaussian window used in S-transform. Simulation results are presented and show correct time frequency information of multiple Gaussian echoes under low signal-to-noise ratio (SNR) environment. In addition, experimental results demonstrate better and reliable detection of close echoes drowned in the noise.     

说话人识别使用遗传RBF网络

针对RBF网络普遍采用的一种训练算法所存在的局部最佳问题,本文将遗传算法用于RBF网络训练过程,增强了该网络的全局寻优能力,提高了RBF网络的模式识别性能.说话人识别实验显示,改进训练算法后,RBF网络的说话人识别率有稍许提高.     

Ultrasonic flaw detection in model railway wheels

Rayleigh and Lamb waves have been excited in $\text{1}\text{4}$ scale model railway wheels by an ultrasonic source located on the wheel tread. Simulated flaws in treads and plates have been located.     

一种新型广义RBF神经网络在混沌时间序列预测中的研究

提出了一种新颖的广义径向基函数神经网络模型,其径向基函数(RBF)的形式由生成函数确定.然后,给出了易实现的梯度学习算法,同时为了进一步提高网络的收敛速度和网络性能,又给出了基于卡尔曼滤波的动态学习算法.为了验证网络的学习性能，采用基于卡尔曼滤波算法的新型广义RBF网络预测模型对Mackey-Glass混沌时间序列和Henon映射进行了仿真.结果表明,所提出的新型广义RBF神经网络模型能快速、精确地预测混沌时间序列,是研究复杂非线性动力系统辨识和控制的一种有效方法. 关键词： 广义径向基函数神经网络 卡尔曼滤波 梯度下降学习算法 混沌时间序列 预测     

Ultrasonic flaw detection in NDE of highly scattering materials using wavelet and Wigner-Ville transform processing

In ultrasonic non-destructive evaluation of highly scattering materials the backscattering noise may attain peak values greater than the searched flaw pulse and the mean value of noise spectrum is very similar to the searched echo spectrum. Several specific methods have been proposed for the reduction of this type of noise, but the comparison of the performance of different methods is still an open problem. In this paper, we make a comparison among some methods based on simultaneous representations in time and frequency/scale domains of the ultrasonic traces. Synthetic and experimental traces are de-noised using a discrete wavelet processor with decomposition level-dependent threshold selection and a method that combines Wigner-Ville transform and filtering in the time-frequency domain. The results are comparatively evaluated in terms of signal to noise ratio and probability of detection.     

基于径向基函数网络的激光诱导荧光特征光谱分离算法

应用激光诱导荧光技术测量水中溶解有机物(DOM)含量，具有灵敏度高、检测速度快、可遥测等优点，其中特征荧光光谱的分离在系统中占有十分重要的地位.在分析激光诱导荧光光谱特征的基础上，提出了采用径向基函数网络(RBFN)分析荧光光谱数据的数学模型，应用这种模型从荧光光谱中恢复出了激光、拉曼和DOM的荧光等光谱分量成分，从而得到了水中DOM的浓度. 关键词： 径向基函数网络 激光诱导荧光 溶解有机物     

Ultrasonic flaw detection in the aircraft industry Aspects of technique and standardization

Aircraft in service are tested in widely different climates, with different equipment and by operators far away from the manufacturer's laboratories and staff. If their inspections are to be valid, the manufacturer must standardize his tests to eliminate all possible sources of error, both technical and human     

基于径向基函数神经网络的混沌时间序列相空间重构双参数联合估计

鉴于径向基函数(RBF)神经网络模型在非线性预测方面的优良性能, 提出了利用该预测模型对混沌时间序列相空间重构的两个关键参数——延迟时间和嵌入维数进行联合估计的方法, 并以客观的评价指标为依据给出其最优估计值. 以Lorenz系统为例进行数值分析, 得到RBF单步及多步预测模型中嵌入维数和延迟时间的最佳参数估计值, 并在原模型中对估计值进行校验. 结果表明, 该方法可以有效地估计出嵌入维数和延迟时间, 从而显著提高预测精度.     

基于混沌理论和改进径向基函数神经网络的网络舆情预测方法

网络舆情发展趋势具有混沌系统的特征, 提出一种基于EMPSO-RBF神经网络的方法对网络舆情的发展趋势进行预测. 首先根据Lyapunov指数证明网络舆情具备混沌的特征, 然后对网络舆情时间序列数据进行相空间重构, 最后采用EMPSO-RBF方法进行预测, 并和其他模型进行对比试验, 实验结果表明EMPSO-RBF方法具有较高精确度.     

Ultrasonic elastography using sector scan imaging and a radial compression

Elastography is an imaging technique based on strain estimation in soft tissues under quasi-static compression. The stress is usually created by a compression plate, and the target is imaged by an ultrasonic linear array. This configuration is used for breast elastography, and has been investigated both theoretically and experimentally. Phenomena such as strain decay with tissue depth and strain concentrations have been reported. However in some in vivo situations, like prostate or blood vessels imaging, this set-up cannot be used. We propose a device to acquire in vivo elastograms of the prostate. The compression is applied by inflating a balloon that covers a transrectal sector probe. The 1D algorithm used to calculate the radial strain fails if the center of the imaging probe does not correspond to the center of the compressor. Therefore, experimental elastograms are calculated with a 2D algorithm that accounts for tangential displacements of the tissue. In this article, in order to gain a better understanding of the image formation process, the use of ultrasonic sector scans to image the radial compression of a target is investigated. Elastograms of homogeneous phantoms are presented, and compared with simulated images. Both show a strain decay with tissue depth. Then experimental and simulated elastograms of a phantom that contains a hard inclusion are presented, showing that strain concentrations occur as well. A method to compensate for strain decay and therefore to increase the contrast of the strain elastograms is proposed. It is expected that such information will help to interpret and possibly improve the elastograms obtained via radial compression.     

Ability of the radial basis function approach to extrapolate nuclear mass

The ability of the radial basis function(RBF) approach to extrapolate the masses of nuclei in neutron-rich and superheavy regions is investigated in combination with the Duflo-Zuker(DZ31), Hartree–Fock-Bogoliubov(HFB27), finite-range droplet model(FRDM12) and Weizs?cker-Skyrme(WS4) mass models. It is found that when the RBF approach is employed with a simple linear basis function, different mass models have different performances in extrapolating nuclear masses in the same region, and a single mass model may have different performances when it is used to extrapolate nuclear masses in different regions. The WS4 and FRDM12 models(two macroscopic–microscopic mass models), combined with the RBF approach, may perform better when extrapolating the nuclear mass in the neutron-rich and superheavy regions.     

基于径向基神经网络预测的混沌时间序列嵌入维数估计方法

根据Takens定理,研究了混沌时间序列相空间重构嵌入维数的选取问题.提出了基于径向基函数神经网络预测模型性能的嵌入维数估计方法,即根据嵌入维数与混沌时间序列预测模型性能的变化关系来确定嵌入维数.通过对几种典型混沌动力学系统的数值验证,结果表明该方法能够确定出合适的相空间重构嵌入维数. 关键词： 混沌 相空间重构 嵌入维数 预测     

Strain field estimation based on digital image correlation and radial basis function

Two methods based on digital image correlation (DIC) and radial basis function (RBF) were proposed to obtain the accurate strain field in this paper. One is a combined method. RBF was applied to remove the noisy discrete displacement data first. After that, the strain was computed by a local least-squares algorithm. The other is a partial derivative of RBF (PD-RBF) based strain estimation method which integrated denoising with differential process. The effectiveness and accuracy of the proposed methods were verified through two numerical simulation experiments. A practical application on the normal strain measurement of an aluminum alloy beam under symmetric four-point bending via an outer loading frame was also presented. The measurement results are in good accordance with the data obtained by strain gauges. Furthermore, a shape parameter selection method based on rate of convergence was suggested. The new method simplifies the choice of the good shape parameter.     

Diffuseness effect and radial basis function network for optimizing a decay calculations

A radial basis function network(RBFN)approach is adopted for the first time to optimize the calcula-tion of a decay half-life in the generalized liquid drop mod...     

Efficient mesh motion using radial basis functions with data reduction algorithms

Mesh motion using radial basis functions has been demonstrated previously by the authors to produce high quality meshes suitable for use within unsteady and aeroelastic CFD codes. In the aeroelastic case the structural mesh may be used as the set of control points governing the deformation, which is efficient since the structural mesh is usually small. However, as a stand alone mesh motion tool, where the surface mesh points control the motion, radial basis functions may be restricted by the size of the surface mesh, as an update of a single volume point depends on all surface points. In this paper a method is presented that allows an arbitrary deformation to be represented to within a desired tolerance by using a significantly reduced set of surface points intelligently identified in a fashion that minimises the error in the interpolated surface. This method may be used on much larger cases and is successfully demonstrated here for a 10⁶ cell mesh, where the initial solve phase cost reduces by a factor of eight with the new scheme and the mesh update by a factor of 55. It has also been shown that the number of surface points required to represent the surface is only geometry dependent (i.e. grid size independent), and so this reduction factor actually increases for larger meshes.