Abstract—Ultrasonic nondestructive testing has been widely used in the detection and evaluation of fatigue cracks and defects of high-speed railway. In order to improve the detection speed of cylindrical components such as rims, a theoretical model of the forward vector algorithm of ultrasonic immersion curved surface is established. For improving image quality, an averaging correction factor is introduced in this paper to avoid multiple irradiation of the same pixel by the same element, and the visualization of cylindrical components with internal defects is realized by simulation and experiment. The traditional synthetic aperture focusing technique (SAFT), Newton iterative synthetic aperture (NISA) and ultrasonic forward vector algorithm (UFVA) are compared and analyzed on the aspects of imaging effect, array performance index (API) and imaging efficiency. Compared with the other two algorithms, the ultrasonic forward vector algorithm can effectively overcome the iteration of calculation of the refraction point and improve the imaging efficiency. The efficiency is improved by 2 and 1 orders of magnitude respectively while imaging quality is guaranteed. At the same time, this algorithm plays an important role on the measurement of the inner and outer diameter of the rim, the defect detection, and the quantitative analysis of the surface wear, which provides an effective means for the rapid detection of internal defects of the regular curved parts.

This study proposes a multi-scale key point extraction algorithm based on normal weighting to address the sensitivity to noise and dependency on object models' shape features in traditional keypoint detection algorithms. First, at each scale, the covariance matrix of the local neighborhood is established and the ratio of the local coordinate system appearing on the first two axes is calculated. Thus, candidate keypoints arc determined based on the ratio. Then, to measure the local maximum dissimilarity measured value of the point cloud, the normal weighted shape index value is calculated. Finally, the maximum value point of the local maximum dissimilarity measured value at different scales is selected as the final keypoint. The experimental results show that compared with other traditional algorithms, the proposed algorithm can effectively extract keypoints of various point cloud models and simultaneously consider the quality and quantity of keypoints and operating efficiency. Moreover, the proposed algorithm has strong adaptability for models with sharp features and large area smooth features, which enhances its robustness and shape index function.

随着铁路向高速、重载及高密度运行的方向发展,钢轨磨耗及表面缺陷日益加剧.基于复合光栅的相位测量轮廓术(phase measurement profilometry,PMP)在保有传统PMP高精度优点的基础上,缩短了测量时间,提高了测量效率.以钢轨磨耗及表面缺陷的三维测量为目的,将传统PMP及基于复合光栅的PMP进行了仿真实现、实验验证与对比分析.同时,在铁路运用背景下,对基于复合光栅的PMP算法进行改进,衍生出基于双频率通道和2+1算法的复合光栅PMP,减小了频谱混叠,提高了测量精度.理论仿真及实验结果表明,改进后的方法测量精度有了明显的提升,且能较好的重建钢轨的三维形貌和表面缺陷,相较于传统复合光栅,测量精度提升了54.9％.

以反向传播神经网络为基础,引入改进的蝙蝠算法对其初始阈值和权值进行优化处理,并针对滚动轴承信号的特征针对性构建了故障诊断系统.针对轴承数据的振动信号选取时、频、多尺度排列熵等提取方式进行多特征参量提取,构造了滚动轴承正常及故障状态下的特征样本并对优化后的神经网络进行训练.然后,使用训练完成的网络对各状态下的随机样本进行诊断测试,诊断结果表明,本文构建的神经网络系统与未优化的BP神经网络相比,可以更为准确地识别出滚动轴承的故障类型,误差降低约一个量级,与未改进的优化算法相比,所介绍的改进算法在保证精度的同时可以有效增加算法的优化效率,同时对强噪音环境下的缺陷具有更高的鉴别率,更高的实用价值.

The 3D point cloud data obtained from the laser line structured light scanner has redundancy, and a point cloud simplification algorithm based on the two order non-uniform partition is designed and implemented to deal with locomotive running department in this paper. First, according to the intrinsic shape signature (ISS), the point cloud normal vector of the detected object are estimated and the feature points of the point cloud are extracted. Then, according to the distribution of the feature point cloud, the point cloud is first divided non-uniformly to obtain uneven initial cloud patches. Finally, the divided cloud points are mapped to different Gaussian spheres for further subdivision. The mean shift clustering is performed on the Gauss sphere to extract the center of gravity of each cluster in the actual three-dimensional space. The set of the center of gravity is the result of simplification. Experimental results verified the effectiveness of the proposed method. It can keep the details information of the point cloud while ensuring a high simplification rate. Comparing with the existing method, this method balances the speed and accuracy, and is more suitable for the on-line locomotive automated detection system.

In this paper, a method of feature extraction in three-dimensional data obtained by laser line structure light is investigated and implemented, for extracting the feature of key parts of locomotive bottom and the locomotive bolt is taken as an example for experimental testing. We use the eigenvalues of the covariance matrix as the features to cluster a number of ribbons, and then use the ISS (Intrinsic Shape Signatures) -based method to get the key points of data in each cluster. The key point is projected to the local surface which is fitted by the least square method based on the key point to form a smooth feature line. The results show that this method is effective, and the feature extraction of 3D point cloud image based on cluster analysis is feasible in railway environment.

In order to ensure the position of the refraction point of the double-layer medium in the full-focus imaging algorithm when in immersion test, this paper carries out the time of flight trajectory of the research, put forward by the method of extreme points to look for refraction and put forward the ray intersection method as a Newton iterative method of the initial test point. The theory of this method is deduced to satisfy Snell's law of refraction and Fermat principle. A 64 array full matrix acquisition and full focus imaging algorithm is used to verify the computational efficiency and imaging quality of the extreme iteration method (EIM). The results show that the imaging speed of the EIM is about 50 times faster than that of the analytical solution model.

针对极限学习机隐含层节点数需人为设定,分类的准确性与稳定性较差,核极限学习机(K-ELM)对核函数选取要求较高,单一核函数难以对非线性样本充分学习、泛化性仍有不足等缺点,提出一种基于多尺度排列熵(MPE)和非线性加权组合的双核极限学习机(DK-ELM)的滚动轴承故障诊断方法并证明了其可行性与优越性.首先,计算不同故障状态轴承信号的多尺度排列熵,获取一系列无量纲特征;然后,利用双核函数计算其高维特征向量集并输入DK-ELM中建立轴承信号状态分类模型,对不同状态的轴承信号进行分类.实验结果证明,核函数的引入可以有效提高ELM分类性能,DK-ELM的分类模型比支持向量机(SVM)、ELM以及各单核极限学习机具有更高的分类精度,而且对训练样本数量较少的情况有更好的分类效果.

Image segmentation is the most fundamental part of computer vision, which is the foundation of all other methods of image processing. The quality of image segmentation technology will affect the subsequent processing considerably. Comparing with traditional image segmentation algorithms, image segmentation algorithm based on deep learning is constantly proposed, with high performance and efficiency. But there is also a lot of room for improvement. For example, key parts such as fastening bolt are usually small in size, polluted and covered, and do not have enough characteristic information, so it is difficult to obtain satisfactory results. These factors affect the accuracy of the test, which is easy to cause serious accidents. As traditional methods sometimes cannot meet the requirement of high-accuracy result, deep learning play a particularly important role in facing those problems. To solve the problem that traditional object recognition methods are not robust enough to extract image features, parts recognition accuracy is low, and segmentation is not possible, we have made some modifications based on Mask R-CNN. In this method, convolutional neural network is used to extract features from part images. Then we use some annotated images from dataset to fine-tuned Mask R-CNN network to guarantee the accuracy. At the same time, data enhancement and k-folding cross-validation are carried out to improve the robustness of the model. Finally, the result of part recognition and segmentation by building the experimental platform proves the significance of the method.

本文引入结构光投影,通过投影随机散斑的方式,增加目标物体的特征点.实验结果表明,投影能有效的增加匹配点数、提高匹配精度.将该方法用于列车底部件的三维成像,得到了较好的三维成像结果.

A two-pump fiber optical parametric amplifier (FOPA) based on the photonic crystal fiber (PCF) with As2S3 background in the mid-infrared (MIR) region is investigated numerically. The genetic algorithms are used to optimize the gain of the FOPA, and the amplifier peak gain, bandwidth and flatness are investigated in detail for the variety of the fiber length, the input signal and pump power. In addition, the comparison of the gain spectra between considering and neglecting the loss of the PCF is given. The results show that the wideband gain spectra with high peak gain can be obtained by using a short length of the PCF with the relatively low pump power, which show the great potential of the FOPA at wavelengths in the MIR region.

针对高速列车重要部件之一的空心轴内部斜裂纹和垂直裂纹难以检测的问题,本文研究了一种双层介质下超声合成孔径空心轴缺陷检测技术.通过费马原理和波束序列合成技术建立了双层介质下相控阵单元超声发射和接收的数值模型,并进行了实验验证,实现了对内部缺陷进行精确定量检测.其结果与直接耦合情况相对比,发现该算法还能有效抑制伪影现象并显著提高信噪比,更具有优越性.

A two-pump fiber optical parametric amplifier (FOPA) based on the photonic crystal fiber (PCF) in the telecommunication region is investigated numerically. The fiber loss and pump depletion are considered. The influences of the fiber length, input signal power, input pump power, and the center pump wavelength on the gain bandwidth, flatness, and peak gain are discussed. The 6-wave model-based analysis of two-pump FOPA is also achieved and compared with that based on the 4-wave model; furthermore, the gain properties of the FOPA based on the 6-wave model are optimized and investigated. The comparison results show that the PCF-based two-pump FOPA achieves flatter and wider gain spectra with less fiber length and input pump power compared to the two-pump FOPA based on the normal highly nonlinear fiber, where the obtained results show the great potential of the FOPA for the optical communication system.

With the rapid arising of information technology, microwave absorbing materials (MAMs) are playing an increasingly significant role in electronic reliability, healthcare, and national defense security. Hence, development of high performance MAMs with thin thickness, low density, wide bandwidth, and strong absorption has attracted great interests. Recently, taking graphene as MAMs for high-performance electromagnetic (EM) wave attenuation has grabbed considerable attention, owing to their low density, high specific surface area, strong dielectric loss, and high electronic conductivity. Furthermore, in order to address the interfacial impedance mismatching of the sole graphene materials, incorporation of other lossy materials has been widely studied as the imperative solution to improve its MA performance. In this review, we introduce the theory of microwave absorption and summarize recent advances in the fabrication of graphene-based MAMs, including rational design of the microstructure of pure graphene and tunable chemical integrations with polymers, magnetic metals, ferrites, ceramics, and multicomponents composites. The key point of enhancing MA in graphene-based MAMs is to regulate their EM properties, improve of impedance matching, and create diversified loss mechanisms. Furthermore, the shortcomings, challenges, and prospects of graphene-based MAMs are also put forward, which will be helpful to people working in the related fields.

In view of their super capacity of adsorbing microbial, the carbon nanotubes (CNTs) were used as the carriers for in situ synthesizing well-dispersed and small-sized silver nanodots (AgND), to prepare a new type of antibacterial agent with remarkably improved activity. Polyethyleneimine (PEI) was introduced as a linkage for guaranteeing the as-generated AgND to be anchored onto the CNTs and to prevent them from agglomeration. The obtained hybridizing materials were characterized by transmission electron microscopy, X-ray diffraction and X-ray photoelectron spectroscopy, and the results showed that the AgND with an average size of 2.6 nm were uniformly loaded on surfaces of CNTs. There existed special interactions between silver atoms and CNTs. The antibacterial activities of the as-prepared hybrids against Escherichia coli were evaluated by disk diffusion assay method and minimal inhibitory concentration measurements. The results showed that the as-prepared hybrids displayed a remarkable improvement in antibacterial activity as compared to CNTs, acidified-CNTs and even the identical silver amount of AgNO3 solution, which was mainly attributed to the small size of AgND and the hybridizing effect between AgND and CNTs.

To overcome the problem of long imaging time caused by the calculation of the refraction point in the double-layer medium ultrasonic imaging, the algorithm in this paper based on the radial calculation of the ultrasonic array element and the radiation arc spline compensation is adopted. And the water-immersed ultrasound total focusing phased-array imaging is discussed. First, according to the half-power beam angle corresponding to the ultrasonic transducer array element, the dividing line of the refraction point is divided, and then Bresenham’s line scanning technology is used to determine all the pixels on the refraction line, and further the delay rule is adopted to determine the pixel point in the transmission process, and finally the reconstructed image is generated. This method not only avoids the large amount of time it takes to iteratively solve the refraction point, but also applies it to the full capture technique and the total focusing imaging, which can save the calculating time of the refraction point to the utmost, thus effectively improving the detection efficiency and imaging quality. Through experimental verification, the efficiency of this method is improved by two orders of magnitude compared to the reverse iteration method in 64-element immersed ultrasound total focusing imaging, and the imaging quality is improved slightly.

三维点云数据的多视角拼接技术是三维测量领域的研究重点之一.文中设计实现了一种使用精简数据引导的迭代最近点(ICP)算法来实现对点云数据快速、高效的拼接.为了验证算法的有效性,本文整理实现了几种常见的快速拼接方案,使用bunny数据对它们进行了拼接对比实验.用作对比的方法包括使用K-d树加速的ICP算法,以及基于特征提取与匹配的拼接方法.实验结果表明,使用精简数据引导的拼接方法运算速度最快,效率最高,同时保持了与其它方法相近的拼接精度.

为解决合成聚焦技术用于超声成像中存在数据存储量大的问题,针对超声回波信号在频域内存在稀疏性的特点,利用压缩感知技术对其在频域进行压缩采样,使用最优化方法完成对回波信号的重建,最终将恢复的信号送入成像系统完成超声成像.通过仿真实验,将该方法用于点目标进行非均匀稀疏采样,所获得的压缩信号恢复出原始信号,完成了成像.结果表明,基于压缩感知的合成聚焦成像方法,在保证成像质量的前提下,可以大幅度减少数据量,降低系统复杂度,仅使用50％数据量,图像质量仍未出现明显失真.

为了保障高速列车安全运行,采用ARIMA数据挖掘理论,研究了高速列车运动部件车轮的直径和磨耗的变化规律和安全预测.实验结果表明,ARIMA模型适用于车轮尺寸预测,随着输入数据量的增加,预测结果越好,而列车轮径比踏面磨耗的预测精度更高;同时与Least Square Estimation的预测结果进行了比较,结果优于Least Square Estimation.这为进一步将ARIMA应用于铁道领域列车关键部件的寿命和安全预测的研究提供参考.

The locomotive running gear 3D point cloud data are obtained by line-structured laser scanner, and the bolts on the locomotive running gear under the 3D point cloud data are recognized and located automatically. Firstly, fast point feature histograms (FPFHs) of the key points are calculated to describe the 3D features, and the target region is matched with the preselected bolt template. Then, K-means clustering is carried out on the weight-ed match point set using uniform seed points. Finally, the Hough transform method is used to establish a strict clas-sifier for the clusters, and the existence and precise position of the bolts are determined. The experimental results verify the effectiveness of the proposed method.