混凝土结构缺陷的融合识别研究

混凝土结构缺陷的无损检测是一项非常困难的工作,尤其是小尺寸或浅层的缺陷检测非常困难,由于仅用一种检测方法往往难以给出令人信服的结论,因此,本文同时采用超声检测和脉冲回波检测两种方法对不同大小的剥离和空洞缺陷进行了探测,并用小波分析方法对这两种信号进行了特征抽取,以第三阶尺度上的极大模作为信号的特征微量;随后,用一个多层前馈神经网络进行了单种检测方法的软决策,定义并计算了以该软决策为基础的概率分配函数;最后,用证据理论方法进行了两种检测方法的决策级融合识别,分类试验结果表明融合识别确实好于单一方法的识别。     

Automatic Defect Classification in Ultrasonic NDT Using Artificial Intelligence

A methodology is developed to detect defects in NDT of materials using an Artificial Neural Network and signal processing technique. This technique is proposed to improve the sensibility of flaw detection and to classify defects in Ultrasonic testing. Wavelet transform is used to derive a feature vector which contains two-dimensional information on various types of defects. These vectors are then classified using an ANN trained with the back propagation algorithm. The inputs of the ANN are the features extracted from each ultrasonic oscillogram. Four different types of defect are considered namely porosity, lack of fusion, and tungsten inclusion and non defect. The training of the ANN uses supervised learning mechanism and therefore each input has the respective desired output. The available dataset is randomly split into a training subset (to update the weight values) and a validation subset. With the wavelet features and ANN, good classification at the rate of 94% is obtained. According to the results, the algorithms developed and applied to ultrasonic signals are highly reliable and precise for online quality monitoring.     

Ultrasonic Defect Mapping Using Signal Correlation for Nondestructive Evaluation (NDE)

This article presents the application of a signal correlation technique to automatically classify ultrasonic A-scan signals for defect and defect-free regions in isotropic and anisotropic materials. First, feature extraction was implemented by generating a reference A-scan signal of a defect-free area using an autocorrelation function and statistics. Then, a cross-correlation function, utilized as a feature detector, was applied to the reference signal and a signal of interest (SOI) to detect defect-free features in an SOI. The correlation result was considered as a pattern containing both defect and defect-free features. Next, the pattern was classified by measuring the similarity between features of the reference signal and an SOI based on their Euclidean distance. Each A-scan signal classification result was then plotted on a 2D map based on its position on the specimen. The present work uses multiple correlation functions and statistics to classify defect signals rather than relying on an inspector’s prior knowledge to interpret C-scan data, and has particular value in automated ultrasonic signal classification and characterization.     

A Guided Wave Approach to Defect Detection in Switch Rail

An approach to detect switch rail defect based on ultrasonic guided wave technology is studied. Eight typical cross-sections are chosen from the switch rail, and each cross-section's theoretical dispersion curves and wave structures are derived using the semianalytical finite-element method. The dispersion curve of eight sections has 262 modes at 60 kHz. Based on k-means clustering analysis algorithm, 25 kinds of classification results are obtained from 262 modes. According to the mean and variance of energy, an optimal set of modes with uniform amplitude over the entire section of the switch rail is selected. The optimal excitation point is determined based on the vibration energy. The defect echo signal is obtained by excitation of these guided wave modes. Field experiment results show that by comparing the sum of the differences between the collected and basic data and setting a certain threshold, the internal defect detection of the switch guide rail can be realized. The research results herein are valuable for analyzing the dispersion characteristics and realizing the nondestructive testing of the variable cross-section waveguide.     

Classification of audio radar signals using radial basis function neural networks

Radial basis function (RBF) neural networks are used to classify real-life audio radar signals that are collected by a ground surveillance radar mounted on a tank. Currently, a human operator is required to operate the radar system to discern among signals bouncing off tanks, vehicles, planes, and so on. The objective of this project is to investigate the possibility of using a neural network to perform this target recognition task, with the aim of reducing the number of personnel required in a tank. Different signal classification methods in the neural net literature are considered. The first method employs a linear autoregressive (AR) model to extract linear features of the audio data, and then perform classification on these features, i.e, the AR coefficients. AR coefficient estimations based on least squares and higher order statistics are considered in this study. The second approach uses nonlinear predictors to model the audio data and then classifies the signals according to the prediction errors. The real-life audio radar data set used here was collected by an AN/PPS-15 ground surveillance radar and consists of 13 different target classes, which include men marching, a man walking, airplanes, a man crawling, and boats, etc. It is found that each classification method has some classes which are difficult to classify. Overall, the AR feature extraction approach is most effective and has a correct classification rate of 88% for the training data and 67% for data not used for training.     

超声C扫描设备定量评价方法研究

超声C扫描系统在超声成像检测、缺陷识别等无损检测领域获得了广泛应用。但是,对C扫描图像的缺陷进行精确分析和表征一直是超声领域的难点之一。基于超声C扫描缺陷图像,给出了一种结合K-means聚类与Graham算法的图像特征参数定量估计方法,通过定量估计的参数能够有效地评价超声C扫描系统的检测质量。实验结果表明该方法能够有效描述标准圆形人工缺陷区域特征,有利于进一步评价超声C扫描设备。     

Identification of feature set for effective tool condition monitoring by acoustic emission sensing

In tool condition monitoring systems, various features from suitably processed acoustic emission signals are utilized by researchers. However, not all of these features are equally informative in a specific monitoring system: certain features may correspond to noise, not information; others may be correlated or not relevant for the task to be realized. This study comprehensively takes all these known signal features and aims to identify the most effective set that can give robust and reliable identification of tool condition. In this paper, the aim is investigated through feature selection, in which automatic relevance determination (ARD) under a Bayesian framework and support vector machine (SVM) are coupled together to perform this task. In tool condition monitoring, this proposed method is able to identify the worst features according to their corresponding ARD parameters and delete them. Then the effectiveness of this pruning may be evaluated by a model validation. Finally, the effective feature set in the developed tool wear recognition system is obtained. The experimental results show that the AE feature set selected through this method is more effective and efficient to recognize tool status over various cutting conditions.     

Neural classification of Lamb wave ultrasonic weld testing signalsusing wavelet coefficients

This paper presents an ultrasonic nondestructive weld testing method based on the wavelet transform (WT) of inspection signals and their classification by a neural network (NN). The use of Lamb waves generated by an electromagnetic acoustic transducer (EMAT) as a probe allows us to test metallic welds. In this work, the case of an aluminum weld is treated. The feature extraction is made by using a method of analysis based on the WT of the ultrasonic testing signals; a classification process of the features based on a neural classifier to interpret the results in terms of weld quality concludes the process. The aim of this complete process of analysis and classification of the testing ultrasonic signals is to lead to an automated system of weld or structure testing. Results of real-world ultrasonic Lamb wave signal analysis and classifications for an aluminum weld are presented; these demonstrate the feasibility and efficiency of the proposed method     

水利工程中金属焊缝缺陷的超声检测方法比较研究

为研究基于超声的无损探伤方法在水利工程金属结构焊缝缺陷识别中的应用,利用常规超声检测技术、超声相控阵技术、衍射时差法(Time of Flight Diffraction, TOFD)超声检测技术对水利工程金属结构焊接试块缺陷进行识别,分析了各种缺陷在超声无损探伤技术中的特征显示。研究结果表明:常规超声检测技术、TOFD检测技术均能对各种缺陷实现信号显示,超声相控阵检测技术对气孔和横向裂纹的显示不够明显,但对其它缺陷的检出效果较为明显;常规超声检测技术对操作人员的要求较高,对缺陷的定性困难,精度不高;TOFD检测结果中气孔和横向裂纹的显示呈现出一种特殊的弧形,有一定高度的内部裂纹和未熔合的信号由上下尖端衍射波组成,根部未焊透上下尖端信号不够明显;相控阵检测结果直观,可以较精确地测量缺陷的埋藏深度、自身高度、长度等,但在扫查点状缺陷或者与超声声束平行的裂纹缺陷时,检出率极低。     

Feature Extraction for Crack Detection in Rain Conditions

The solution of the features selection problem is critical for robust detection of crack signals in noisy environment, varying from short-time impulses such as raindrops to the wide-band white Gaussian noise. In this paper, two novel feature selection methods were used to reduce an initial set of 90 features, 67 estimated in the time domain and 23 in the frequency domain, decreasing significantly the memory requirements and the computational complexity of a Radial-Basis-Function (RBF) cracks detector. The evaluation process is carried out in a database including of more than 6000 cracks, raindrops and simultaneous crack and raindrops signals. Additive white Gaussian noise is used to distort the real signals at ?20 to 20 dB Signal to Noise Ratio (SNR). The experimental results show that the number of features can be reduced to approximately 25, without affecting the classification rate of cracks and raindrops in the noisy signals, if the SNR is better than 0 dB. In noise-free environment a classification rate of 91% for a single crack/raindrop event is achieved using only five features. A different set of five features reaches a rate of 85% at 10 dB SNR.     

一种基于Triplet loss的齿轮箱复合故障识别方法

随着设备检测点的数量与采样频率的增加,机械健康监测进入了"大数据"时代。深度学习以其强大的自适应特征提取和分类能力也在机械大数据处理方面取得了丰硕的成果。在故障诊断领域,目前深度学习方法的研究对象均集中于单一故障,而复合故障却鲜有人涉足。复合故障因为其各类故障信号间有耦合,变化的工况(负载,转速)也会对信号产生较大影响,所以难以准确诊断。面对复杂的复合故障,传统的Softmax分类器已不能精确高效的完成故障诊断。提出了一种基于Triplet loss的深度度量学习模型的诊断方法,对齿轮箱的轴承及齿轮这两种目标的故障同时进行诊断。其优势在于通过该模型提取故障信号的特征,再利用Triplet loss度量各类故障之间的距离,使得同类故障特征间的距离很近,异类故障特征间的距离很远,从而高效完成诊断任务。试验结果表明,该方法实现了在多种工况,大量样本下对齿轮箱内轴承和齿轮不同故障的准确诊断。     

基于LFM的橡胶脱粘超声检测方法研究

针对SYS510e型空气弹簧底板的金属橡胶粘接结构橡胶脱粘缺陷超声检测难以辨识问题,提出采用改进的线性调频脉冲代替传统窄脉冲作为超声波激励信号,增大超声检测的信号能量和频谱宽度.在宽频带超声检测的基础上,采用小波包-奇异值分解方法解析超声回波在不同粘接状态、不同频率范围的时频能量分布,提取更稳定、一致性更好的橡胶脱粘辨...     

PCA-based feature selection scheme for machine defect classification

The sensitivity of various features that are characteristic of a machine defect may vary considerably under different operating conditions. Hence it is critical to devise a systematic feature selection scheme that provides guidance on choosing the most representative features for defect classification. This paper presents a feature selection scheme based on the principal component analysis (PCA) method. The effectiveness of the scheme was verified experimentally on a bearing test bed, using both supervised and unsupervised defect classification approaches. The objective of the study was to identify the severity level of bearing defects, where no a priori knowledge on the defect conditions was available. The proposed scheme has shown to provide more accurate defect classification with fewer feature inputs than using all features initially considered relevant. The result confirms its utility as an effective tool for machine health assessment.     

基于深度学习的工业自动化包装缺陷检测方法

目的针对目前工业自动化生产中基于人工特征提取的包装缺陷检测方法复杂、专业知识要求高、通用性差、在多目标和复杂背景下难以应用等问题,研究基于深度学习的实时包装缺陷检测方法。方法在样本数据较少的情况下,提出一种基于深度学习的Inception-V3图像分类算法和YOLO-V3目标检测算法相结合的缺陷检测方法,并设计完整的基于计算机视觉的在线包装缺陷检测系统。结果实验结果显示,该方法的识别准确率为99.49%,方差为0.0000506,只使用Inception-V3算法的准确率为97.70%,方差为0.000251。结论相比一般基于人工特征提取的包装缺陷检测方法,避免了复杂的特征提取过程。相比只应用图像分类算法进行包装缺陷检测,该方法在包装缺陷区域占比较小的情况下能较明显地提高包装缺陷检测精度和稳定性,在复杂检测背景和多目标场景中体现优势。该缺陷检测系统和检测方法可以很容易地迁移到其他类似在线检测问题上。     

An automated methodology for the classification of focal and nonfocal EEG signals using a hybrid classification approach

The uncertainty in human brain leads to the formation of epilepsy disease in human. The automatic detection and severity analysis of epilepsy disease is proposed in this article using a hybrid classification algorithm. The proposed method consists of decomposition stage, feature extraction, and classification stages. The electroencephalogram (EEG) signals are decomposed using dual-tree complex wavelet transform and then features are extracted from these coefficients. These features are then classified using the neural network classification approach in order to classify the EEG signals into either focal or nonfocal EEG signals. Furthermore, severity of the focal EEG signal is analyzed using an adaptive neuro-fuzzy inference system classification approach. The proposed hybrid classification method for the classification of focal signals and nonfocal signals achieved 98.6% of sensitivity, 99.1% of specificity, and 99.4% of accuracy. The average detection rate for both focal and nonfocal dataset is about 98.5%.     

Enhanced Deep Autoencoder Based Feature Representation Learning for Intelligent Intrusion Detection System

In the era of Big data, learning discriminant feature representation from network traffic is identified has as an invariably essential task for improving the detection ability of an intrusion detection system (IDS). Owing to the lack of accurately labeled network traffic data, many unsupervised feature representation learning models have been proposed with state-of-the-art performance. Yet, these models fail to consider the classification error while learning the feature representation. Intuitively, the learnt feature representation may degrade the performance of the classification task. For the first time in the field of intrusion detection, this paper proposes an unsupervised IDS model leveraging the benefits of deep autoencoder (DAE) for learning the robust feature representation and one-class support vector machine (OCSVM) for finding the more compact decision hyperplane for intrusion detection. Specially, the proposed model defines a new unified objective function to minimize the reconstruction and classification error simultaneously. This unique contribution not only enables the model to support joint learning for feature representation and classifier training but also guides to learn the robust feature representation which can improve the discrimination ability of the classifier for intrusion detection. Three set of evaluation experiments are conducted to demonstrate the potential of the proposed model. First, the ablation evaluation on benchmark dataset, NSL-KDD validates the design decision of the proposed model. Next, the performance evaluation on recent intrusion dataset, UNSW-NB15 signifies the stable performance of the proposed model. Finally, the comparative evaluation verifies the efficacy of the proposed model against recently published state-of-the-art methods.     

基于3D和1D多特征融合的语音情感识别算法

针对语音情感识别任务中特征提取单一、分类准确率低等问题,提出一种3D和1D多特征融合的情感识别方法,对特征提取算法进行改进.在3D网络,综合考虑空间特征学习和时间依赖性构造,利用双线性卷积神经网络(Bilinear Convolutional Neural Network,BCNN)提取空间特征,长短期记忆网络(Sho...     

应用卷积神经网络的锂电池极片涂布缺陷分类

目的 针对锂电池极片涂布缺陷种类多,传统方法分类检测精度不高,以及人工依赖性强等问题,提出一种基于卷积神经网络的锂电池极片涂布缺陷自动分类算法。方法 首先对网络结构以及模型参数进行优化,接着在网络中加入跳跃连接结构,将空洞卷积提取到的多尺度特征与高层特征进行融合以获取更多缺陷特征,并采用LeakyReLU(Leaky Rectified Linear Unit)激活函数保留图像中的负值特征信息,最后通过构建的数据集训练模型,实现锂电池极片涂布缺陷的准确分类。结果 实验结果表明,当前方法识别准确率能够达到99.34%,平均检测时间为51ms。结论 改进后的方法能够准确分类出锂电池极片18种涂布缺陷,满足工业生产中实时分类检测的要求。     

基于小波包时频图特征和卷积神经网络的水声信号分类

水下声信号分类是水声学研究的一个重要方向.一个有效的特征提取和分类决策方法对水声信号分类技术至关重要.文章将鱼声、商船辐射噪声和风关噪声三类实测的水声信号在小波包分解的基础上提取时频图特征,并搭建了一个七层结构的卷积神经网络作为分类器.研究结果表明:三种水声信号的小波包时频图特征结合卷积神经网络在不同测试集可达到(98...