基于深度置信网络与信息融合的齿轮故障诊断方法

针对齿轮在复杂运行工况下故障特征提取困难,传统故障诊断方法的识别精度易受人工提取特征的影响,以及单传感器获取信息不全面等问题,提出基于深度置信网络(DBN)与信息融合的齿轮故障诊断方法。通过多传感器信息融合技术对每个传感器采集的振动信号进行数据层融合;利用DBN进行自适应特征提取从而实现故障分类。为了避免因人为选择DBN结构参数,导致模型识别精度下降的问题,利用改进的混合蛙跳算法(ISFLA)对DBN结构参数进行优化。试验表明,与BP神经网络、未经优化的DBN以及单传感器故障诊断相比,该研究提出的信息融合及优化方法具有更高的故障识别精度。     

Online Monitoring of Welding Status Based on a DBN Model During Laser Welding

In this research, an auxiliary illumination visual sensor system, an ultraviolet/visible (UVV) band visual sensor system (with a wavelength less than 780 nm), a spectrometer, and a photodiode are employed to capture insights into the high-power disc laser welding process. The features of the visible optical light signal and the reflected laser light signal are extracted by decomposing the original signal captured by the photodiode via the wavelet packet decomposition (WPD) method. The captured signals of the spectrometer mainly have a wavelength of 400–900 nm, and are divided into 25 sub-bands to extract the spectrum features by statistical methods. The features of the plume and spatters are acquired by images captured by the UVV visual sensor system, and the features of the keyhole are extracted from images captured by the auxiliary illumination visual sensor system. Based on these real-time quantized features of the welding process, a deep belief network (DBN) is established to monitor the welding status. A genetic algorithm is applied to optimize the parameters of the proposed DBN model. The established DBN model shows higher accuracy and robustness in monitoring welding status in comparison with a traditional back-propagation neural network (BPNN) model. The effectiveness and generalization ability of the proposed DBN are validated by three additional experiments with different welding parameters.     

基于改进DBN的回转支承寿命状态识别

为了解决大型回转支承背景噪声大,特征信号微弱,寿命状态难以识别等问题,提出了一种基于改进深度信念网络(Deep Belief Network,DBN)的回转支承寿命状态识别方法。DBN网络拥有强大的深度学习能力,能够有效挖掘回转支承运行状态信息,解决了传统浅层网络过度依赖特征提取效果和识别精度不高的问题。在DBN学习训练中,采用新的优化学习方法FEPCD(Free Energy in Persistent Contrastive Divergence),解决了DBN在长期学习中近似和分类能力下降的问题。然后利用自主研发试验台的试验数据对所提方法的优越性进行验证。将改进的DBN算法与浅层分类算法的识别结果进行比较。结果表明改进DBN网络比原始DBN网络和浅层算法能更精确反映回转支承寿命特征,所提方法具有稳定性和智能性的特点。     

基于PSO改进深度置信网络的滚动轴承故障诊断

针对深度置信网络(Deep Belief Network,DBN)用于轴承故障诊断时,网络层结构调试比较费时等问题,提出一种基于粒子群优化(Particle Swarm Optimization,PSO)的DBN算法,以及基于该算法的轴承故障诊断模型。该模型利用PSO算法优选DBN网络结构,并通过自适应时刻估计法微调模型参数,随后运用具有最优结构的DBN模型直接从原始振动信号中提取低维故障特征,并将其输入到Soft-max分类器中识别轴承的故障模式。该算法与支持向量机、BP神经网络、DBN、堆叠降噪自编码等方法进行对比分析,实验结果表明,PSO改进的DBN算法具有更高的准确率以及更好的鲁棒性。     

融合人脸表情的手语到汉藏双语情感语音转换

针对聋哑人与正常人之间存在的交流障碍问题,提出了一种融合人脸表情的手语到汉藏双语情感语音转换的方法。首先使用深度置信网络模型得到手势图像的特征信息,并通过深度神经网络模型得到人脸信息的表情特征。其次采用支持向量机对手势特征和人脸表情特征分别进行相应模型的训练及分类,根据识别出的手势信息和人脸表情信息分别获得手势文本及相应的情感标签。同时,利用普通话情感训练语料,采用说话人自适应训练方法,实现了一个基于隐Markov模型的情感语音合成系统。最后,利用识别获得的手势文本和情感标签,将手势及人脸表情转换为普通话或藏语的情感语音。客观评测表明,静态手势的识别率为92.8%,在扩充的Cohn-Kanade数据库和日本女性面部表情(Japanese Female Facial Expression,JAFFE)数据库上的人脸表情识别率为94.6%及80.3%。主观评测表明,转换获得的情感语音平均情感主观评定得分4.0分,利用三维情绪模型(Pleasure-Arousal-Dominance,PAD)分别评测人脸表情和合成的情感语音的PAD值,两者具有很高的相似度,表明合成的情感语音能够表达人脸表情的情感。     

DDoS Attack Detection via Multi-Scale Convolutional Neural Network

Distributed Denial-of-Service (DDoS) has caused great damage to the network in the big data environment. Existing methods are characterized by low computational efficiency, high false alarm rate and high false alarm rate. In this paper, we propose a DDoS attack detection method based on network flow grayscale matrix feature via multiscale convolutional neural network (CNN). According to the different characteristics of the attack flow and the normal flow in the IP protocol, the seven-tuple is defined to describe the network flow characteristics and converted into a grayscale feature by binary. Based on the network flow grayscale matrix feature (GMF), the convolution kernel of different spatial scales is used to improve the accuracy of feature segmentation, global features and local features of the network flow are extracted. A DDoS attack classifier based on multi-scale convolution neural network is constructed. Experiments show that compared with correlation methods, this method can improve the robustness of the classifier, reduce the false alarm rate and the missing alarm rate.     

基于深度度量学习的轴承故障诊断方法

针对机械大数据因故障类内离散度和类间相似度较大而导致诊断精度低的问题,提出一种深度度量学习故障诊断方法,采用深度神经网络(Deep Neural Network, DNN)对故障特征进行自适应提取,并利用基于欧氏距离的边际Fisher分析(Marginal Fisher Analysis, MFA)方法进行了优选,在构建的深度度量网络(Deep Metric Network, DMN)顶层特征输出层添加BPNN(Back Propagation Neural Network, BPNN)分类器对网络参数进行微调,并实现故障的分类识别。通过对不同类型和严重程度的轴承故障进行了诊断分析,验证了该方法可以有效地对轴承故障进行高精度诊断,效果优于传统深度信念网络(Deep Belief Network, DBN)故障诊断方法以及常用时域统计特征结合支持向量机(Support Vector Machine, SVM)分类的故障诊断方法。     

变载荷齿轮箱故障信号智能检测方法

针对变载荷激励下齿轮故障信号检测和故障识别的问题,提出一种基于经验模态分解和深度信念网络的变载荷齿轮箱故障信号智能检测方法。首先通过经验模态分解方法将非平稳的加速度振动信号分解成若干平稳的本征模态函数;选出啮合频率及倍频所在的本征模态函数,重构信号,求出重构信号的频谱,作为深度信念网络的输入;深度信念网络通过对输入频谱进行预训练和特征学习,并建立变载荷激励下齿轮故障识别分类模型;最后,用构建好的深度信念网络对测试样本进行故障诊断。试验结果表明,提出的方法能有效地检测和识别变载荷激励下齿轮故障。     

基于变分模态分解与深度信念网络的运动想象分类识别研究

传统人工确定最优时段及最优频段的方法会造成信息遗漏进而导致运动想象识别率的降低, 因此基于脑电信号的运动想象分类研究成为了脑-机接口研究领域的难点问题。针对该问题,变分模态分解和深度信念网络被应用于运动想象分类。对脑电信号进行变分模态分解得到窄带分量,利用希尔伯特变换提取边际谱、特征频带下的瞬时能谱以及时-频联合特征; 特征融合后采用深度信念网络对高维特征降维并实现运动想象模式的识别, 避免了人工确定想象最优时段及最优频段造成的信息遗漏。实验结果表明,利用变分模态分解与深度信念网络自动提取最优时段及最优频段特征的方法有效提升了运动想象识别率。     

深度置信网络迁移学习的行星齿轮箱故障诊断方法

实际工程中行星齿轮箱受工况、运行情况等因素的影响,获取的数据难以满足训练和测试数据独立同分布且训练数据充足的条件,直接影响故障诊断效果。为此,提出一种深度置信网络(Deep Belief Network,DBN)迁移学习的行星齿轮箱故障诊断方法。首先,将辅助标记数据的原始信号频谱作为DBN网络的输入,逐层更新网络的权重和偏置值对输入信号进行分级表达,以获得其分布式特征表达,得到基于辅助标记样本的DBN预模型。再利用少量的目标标记样本微调DBN预模型的网络权重和偏置值,实现DBN网络的权重和偏置值从源域到目标域的迁移以适应新的目标样本识别,最终提高目标域样本故障识别准确率。通过行星齿轮箱故障模拟实验验证了所提方法的可行性和有效性。     

基于卷积神经网络和Transformer网络的鸟声识别

针对传统鸟声识别算法中特征提取方式单一、分类识别准确率低等问题,提出一种结合卷积神经网络和Transformer网络的鸟声识别方法。该方法综合考虑网络局部特征学习和全局上下文依赖性构造,从原始鸟声音频信号中提取短时傅里叶变换(Short Time Fourier Transform,STFT)语谱图特征,将其输入到卷积神经网络(ConvolutionalNeural Network,CNN)中提取局部频谱特征信息,同时提取鸟声信号的对数梅尔特征及一阶差分、二阶差分特征用于合成梅尔频率倒谱系数(Mel Frequency Cepstrum Coefficient,MFCC)混合特征向量,将其输入到Transformer网络中获取全局序列特征信息,最后融合所提取的特征可得到更丰富的鸟声特征参数,通过Softmax分类器得到鸟声识别结果。在Birdsdata和xeno-canto鸟声数据集上进行实验,平均识别准确率分别达到了97.81%和89.47%。实验结果表明该方法相较于其他现有的鸟声识别模型具有更高的识别准确率。     

基于深度学习的舰船辐射噪声多特征融合分类

为了改善分类系统的性能,进一步提高舰船辐射噪声分类的正确率,该文提出了一种基于深度神经网络的多特征融合分类方法。该方法首先提取舰船辐射噪声几种不同的特征,将提取的特征同时用于训练具有多个输入分支的深度神经网络,使网络直接在多种特征参数上进行联合学习,通过神经网络的输入分支和连接层实现特征融合,再对舰船辐射噪声进行分类。为了特征深度学习提取了舰船辐射噪声的频谱特征、梅尔倒谱系数和功率谱特征,并将多特征融合分类方法与在一种特征上进行深度学习分类方法的正确率进行对比。实验结果表明,基于深度学习的多特征融合分类方法可以有效地提高舰船辐射噪声分类的正确率,是一种可行的分类方法。     

大数据下的深度学习研究

给出了大数据和机器学习的子领域——深度学习的概念,阐述了深度学习对获取大数据中的有价值信息的重要作用。描述了大数据下利用图像处理单元(GPU)进行并行运算的深度学习框架,对其中的大规模卷积神经网络(CNN)、大规模深度置信网络(DBN)和大规模递归神经网络(RNN)进行了重点论述。分析了大数据的容量、多样性、速率特征,介绍了大规模数据、多样性数据、高速率数据下的深度学习方法。展望了大数据背景下深度学习的发展前景,指出在不远的将来,大数据与深度学习融合的技术将会在计算机视觉、机器智能等多个领域获得突破性进展。     

Detection and classification of buried dielectric anomalies bymeans of the bispectrum method and neural networks

The development of a neural network-based system for detection and classification of buried landmines is the main focus of this paper. Shape-dependent features are extracted by means of the bispectrum method. These features are then applied to the neural network. A multilayer back-propagation-type neural network is trained and tested on the feature sets extracted from equally spaced radial slices of image windows. Simulation results obtained for two types of targets indicated good detection and classification rates     

多域信息融合结合改进残差密集网络的轴承故障诊断

针对滚动轴承原始时域信号信息单一、深度卷积神经网络提取的特征对信息的传递存在差异等问题,该研究提出了一种多域信息融合与改进残差密集网络的轴承故障诊断方法。为了获取故障的多方面信息,先对原始数据进行多域变换,再将融合信息输入经卷积注意力改进的残差密集网络进行深度学习。经注意力机制改进的网络能够实现对提取特征的重要性区分,提高网络的训练速度、改善识别准确率。试验结果及对比分析表明该算法可以提取较为全面的特征,较传统方法具有更好的识别效果。     

基于变分模态分解与深度卷积神经网络的滚动轴承故障诊断

针对滚动轴承振动信号非平稳、非线性特点以及特征提取困难问题,提出一种基于变分模态分解(VMD)与深度卷积神经网络相结合的特征提取方法并应用于滚动轴承故障诊断。利用VMD将原始振动信号分解得到若干不同频率的限带本征模态分量,通过卷积网络中的多组卷积核自动学习各模态数据的不同特征,保证了特征提取的自适应性、全面性和多样性。在特征提取的基础上,使用全连接神经网络进行故障分类与诊断。将所提方法应用于滚动轴承故障诊断,结果表明,该方法在变工况情况下能够实现滚动轴承故障类别以及损伤程度的精确判定。     

基于模糊神经网络与主成分分析的磨削表面粗糙度在线预测

基于声发射和振动信号提出了一种模糊神经网络和主成分分析的表面粗糙度预测方法,以提高磨削过程中工件表面粗糙度识别的准确性。首先,采集磨削程中声发射与振动信号,提取相关时域特征、频域特征和小波包特征参数,利用主成分分析对特征量进行降维优化;然后,构建表面粗糙度模糊神经网络预测模型,将信号特征量与表面粗糙度作为模糊神经网络的输入和输出;最后,对模型进行训练,并对表面粗糙度预测精度进行验证。实验结果表明:通过主成分分析(PCA)方法对声发射和振动信号特征量进行降维得到5个主成分,以此建立的模糊神经网络表面粗糙度预测模型的效果精度可达到91%以上,与局部线性嵌入和多维标度法降维方法相比,PCA方法降维后的特征所含信息更优,预测准确度更高。     

Sentiment Classification Based on Piecewise Pooling Convolutional Neural Network

Recently, the effectiveness of neural networks, especially convolutional neural networks, has been validated in the field of natural language processing, in which, sentiment classification for online reviews is an important and challenging task. Existing convolutional neural networks extract important features of sentences without local features or the feature sequence. Thus, these models do not perform well, especially for transition sentences. To this end, we propose a Piecewise Pooling Convolutional Neural Network (PPCNN) for sentiment classification. Firstly, with a sentence presented by word vectors, convolution operation is introduced to obtain the convolution feature map vectors. Secondly, these vectors are segmented according to the positions of transition words in sentences. Thirdly, the most significant feature of each local segment is extracted using max pooling mechanism, and then the different aspects of features can be extracted. Specifically, the relative sequence of these features is preserved. Finally, after processed by the dropout algorithm, the softmax classifier is trained for sentiment classification. Experimental results show that the proposed method PPCNN is effective and superior to other baseline methods, especially for datasets with transition sentences.     

A DDoS Attack Information Fusion Method Based on CNN for Multi-Element Data

Traditional distributed denial of service (DDoS) detection methods need a lot of computing resource, and many of them which are based on single element have high missing rate and false alarm rate. In order to solve the problems, this paper proposes a DDoS attack information fusion method based on CNN for multi-element data. Firstly, according to the distribution, concentration and high traffic abruptness of DDoS attacks, this paper defines six features which are respectively obtained from the elements of source IP address, destination IP address, source port, destination port, packet size and the number of IP packets. Then, we propose feature weight calculation algorithm based on principal component analysis to measure the importance of different features in different network environment. The algorithm of weighted multi-element feature fusion proposed in this paper is used to fuse different features, and obtain multi-element fusion feature (MEFF) value. Finally, the DDoS attack information fusion classification model is established by using convolutional neural network and support vector machine respectively based on the MEFF time series. Experimental results show that the information fusion method proposed can effectively fuse multi-element data, reduce the missing rate and total error rate, memory resource consumption, running time, and improve the detection rate.     

DDoS Detection in SDN using Machine Learning Techniques

Software-defined network (SDN) becomes a new revolutionary paradigm in networks because it provides more control and network operation over a network infrastructure. The SDN controller is considered as the operating system of the SDN based network infrastructure, and it is responsible for executing the different network applications and maintaining the network services and functionalities. Despite all its tremendous capabilities, the SDN face many security issues due to the complexity of the SDN architecture. Distributed denial of services (DDoS) is a common attack on SDN due to its centralized architecture, especially at the control layer of the SDN that has a network-wide impact. Machine learning is now widely used for fast detection of these attacks. In this paper, some important feature selection methods for machine learning on DDoS detection are evaluated. The selection of optimal features reflects the classification accuracy of the machine learning techniques and the performance of the SDN controller. A comparative analysis of feature selection and machine learning classifiers is also derived to detect SDN attacks. The experimental results show that the Random forest (RF) classifier trains the more accurate model with 99.97% accuracy using features subset by the Recursive feature elimination (RFE) method.