Fault Current Identification of DC Traction Feeder Based on Optimized VMD and Sample Entropy

A current identification method based on optimized variational mode decomposition(VMD)and sample entropy(SampEn)is proposed in order to solve the problem that the main protection of the urban rail transit DC feeder cannot distinguish between train charging current and remote short circuit current.This method uses the principle of energy difference to optimize the optimal mode decomposition number k of VMD;the optimal VMD for DC feeder current is decomposed into the intrinsic modal function(IMF)of different frequency bands.The sample entropy algorithm is used to perform feature extraction of each IMF,and then the eigenvalues of the intrinsic modal function of each frequency band of the current signal can be obtained.The recognition feature vector is input into the support vector machine model based on Bayesian hyperparameter optimization for training.After a large number of experimental data are verified,it is found that the optimal VMD_SampEn algorithm to identify the train charging current and remote short circuit current is more accurate than other algorithms.Thus,the algorithm based on optimized VMD_SampEn has certain engineering application value in the fault current identification of the DC traction feeder.

Efficient slope reliability analysis under soil spatial variability using maximum entropy distribution with fractional moments

Spatial variability of soil properties imposes a challenge for practical analysis and design in geotechnical engineering.The latter is particularly true for slope stability assessment,where the effects of uncertainty are synthesized in the so-called probability of failure.This probability quantifies the reliability of a slope and its numerical calculation is usually quite involved from a numerical viewpoint.In view of this issue,this paper proposes an approach for failure probability assessment based on Latinized partially stratified sampling and maximum entropy distribution with fractional moments.The spatial variability of geotechnical properties is represented by means of random fields and the Karhunen-Loève expansion.Then,failure probabilities are estimated employing maximum entropy distribution with fractional moments.The application of the proposed approach is examined with two examples:a case study of an undrained slope and a case study of a slope with cross-correlated random fields of strength parameters under a drained slope.The results show that the proposed approach has excellent accuracy and high efficiency,and it can be applied straightforwardly to similar geotechnical engineering problems.

A new approach for epileptic seizure detection: sample entropy based feature extraction and extreme learning machine

The electroencephalogram (EEG) signal plays a key role in the diagnosis of epilepsy. Substantial data is generated by the EEG recordings of ambulatory recording systems, and detection of epileptic activity requires a time-consuming analysis of the complete length of the EEG time series data by a neurology expert. A variety of automatic epilepsy detection systems have been developed during the last ten years. In this paper, we investigate the potential of a recently-proposed statistical measure parameter regarded as Sample Entropy (SampEn), as a method of feature extraction to the task of classifying three different kinds of EEG signals (normal, interictal and ictal) and detecting epileptic seizures. It is known that the value of the SampEn falls suddenly during an epileptic seizure and this fact is utilized in the proposed diagnosis system. Two different kinds of classification models, back-propagation neural network (BPNN) and the recently-developed extreme learning machine (ELM) are tested in this study. Results show that the proposed automatic epilepsy detection system which uses sample entropy (SampEn) as the only input feature, together with extreme learning machine (ELM) classification model, not only achieves high classification accuracy (95.67%) but also very fast speed.

Tool State Detection by Harmonic Wavelet and Sample Entropy

It is a fact that acoustic emission(AE) signals contain potentially valuable information for tool wear and breakage monitoring and detection.However,AE stress waves produced in the cutting zone are distorted by the transmission path and the measurement systems,it is difficult to obtain a reliable result by these raw AE data.It is generally known that the process of tool wear belongs to detect weak singularity signals in strong noise.The objective of this paper is to combine Newland Harmonic wavelet and Richman-Moorman(2000) sample entropy for detecting weak singularity signals embedded in strong signals.First,the raw AE signal is decomposed by harmonic wavelet and transformed into the three-dimensional time-frequency mesh map of the harmonic wavelet,at the same time,the contours of the mesh map with log space is induced.Second,the profile map of the three-dimensional time-frequency mesh map is offered,which corresponds to decomposed level on harmonic wavelets.Final,by computing sample entropy in each level,the weak singularity signal can be easily extracted from strong noise.Machining test was carried out on HL-32 NC turning center.This lathe does not have a tailstock.Tungsten carbide finishing tool was used to turn free machining mild steel.The work material was chosen for ease of machining,allowing for generation of surfaces of varying quality without the use of cutting fluids.In turning experiments,the feasibility for tool condition monitoring is demonstrated by 27 kinds of cutting conditions with the sharp tool and the worn tool,54 group data are sampled by AE.The sample entropy of each level of wavelet decomposed for each one of 54 AE datum is computed,wear tool and shaper tool can be distinguished obviously by the sample entropy value at the 12th level,this is a criterion.The proposed research provides a new theoretical basis and a new engineering application on the tool condition monitoring.

Attention Drawing of Movie Trailers Revealed by Electroencephography Using Sample Entropy

A movie trailer is a common advertising tool in the entertainment industry. Detection of a viewer’s brain responses to a movie trailer can help film producers to tailor a more appealing trailer of a movie. In this study, we acquired electroencephalographic (EEG) signals from subjects when they watched movie trailers (labeled as Movie session), and compared with their resting state session (labeled as Resting session) or when they watch nature scenes (labeled as Nature session). We used Sample Entropy (SampEn) to analyze the EEG signals between different sessions. Results showed that the complexity ratios at Fp1, Fp2 and Fz channels derived from Movie session were significantly lower than that in Resting state or when subjects watched Nature session (p < 0.001). Our results suggest that the brain status can affect the complexity of their EEG. Further, the attraction of attention of a movie trailer can be observed from the change of EEG.

Research on natural language recognition algorithm based on sample entropy

Sample entropy can reflect the change of level of new information in signal sequence as well as the size of the new information. Based on the sample entropy as the features of speech classification, the paper firstly extract the sample entropy of mixed signal, mean and variance to calculate each signal sample entropy, finally uses the K mean clustering to recognize. The simulation results show that： the recognition rate can be increased to 89.2% based on sample entropy.

RFOA优化EEMD阈值和SampEn的水电机组振动信号重构与特征提取

针对EEMD在水电机组振动信号降噪处理中的不足,提出一种基于改进果蝇算法(RFOA)优化EEMD噪声IMF分量阈值的降噪算法。通过EEMD算法将噪声信号分解,得到IMF分量,进而通过相关系数法确定噪声信号与有效信号,利用RFOA确定噪声信号IMF分量阈值;将求得的IMF分量的样本熵当作特征向量输入GRNN算法中,进行振动模式识别。研究结果表明,与小波阈值法、EEMD-GA方法相比,所提算法降噪时信噪比最高,降噪效果最佳。

基于VMD-SampEn-M1DCNN组合模型的钳形电流互感器故障诊断

针对钳形电流互感器故障诊断效率和诊断准确率低的问题,提出一种基于VMD-SampEn-M1DCNN组合模型的钳形电流互感器故障诊断模型。首先,以钳形电流互感器数据为基础,对其进行VMD分解,以建立本征模函数(IMF),并进行IMF分量选择;然后,选取IMF分量分析样本熵,并将其作为互感器特征提取对象的特征值;最后构建了M1DCNN模型,对模型进行样本数据训练和测试。结合实验分析结果,证实VMD-SampEn-M1DCNN模型在训练时间、测试时间和模型测试精准度方面,与传统故障诊断相比都有明显的优势,能进行故障的精准诊断。

基于CEEMD-SE的CNN&LSTM-GRU短期风电功率预测

为进一步提升短期风电功率的预测精度,提出了一种基于互补集合经验模态分解-样本熵(complementary ensemble empirical mode decomposition-sample entropy,CEEMD-SE)的卷积神经网络(convolutional neural network,CNN)和长短期记忆-门控循环单元(longshorttermmemory-gatedrecurrentunit,LSTM-GRU)的短期风电功率预测模型。首先,利用互补集合经验模态分解将原始风电功率序列分解为若干本征模态函数(intrinsic mode function,IMF)分量和一个残差(residual,RES)分量,利用样本熵算法将相近的分量进行重构;其次,搭建卷积神经网络和长短期记忆网络的并行网络结构,提取数据的局部特征和时序特征,并将特征融合后输入门控循环单元网络中进行学习预测;最后,通过算例进行验证,结果表明采用该模型后预测精度得到了有效提升,其均方根误差降低了15.06%、平均绝对误差降低了15.22%、决定系数提高了1.91%。

最大熵试验法对切割器功能可靠性的评估

针对切割器设计裕度大且刺激能量服从正态分布的特点,提出采用最大熵试验法对切割器的功能可靠性进行评估。以某型切割器为实例,通过临界药量摸底试验、升降法试验及成败型试验对其功能可靠性进行了评估。结果表明:运用最大熵试验法,采用46个试验样本即可完成切割器在置信度γ=0.95下,可靠性指标不小于0.999 995的可靠性评估试验,为航天火工装置(尤其是切割器)的小样本量高可靠性评估提供了技术途径。

入侵意图分析下的软件定义网络DDoS攻击检测方法

为在数据样本回溯期内解决因本地信息熵值增大而造成的服务攻击问题,维护软件定义网络的运行安全性,提出入侵意图分析下的软件定义网络分布式拒绝服务(DDoS)攻击检测方法。按照软件定义网络场景重构原则,确定因果网转换标准,实现对识别参数的更新处理,完成攻击性行为的入侵意图分析,再定义DDoS数据集,根据攻击行为的时空特性,求解模型参数的取值范围,完成入侵意图分析下软件定义网络DDoS攻击检测方法的设计。实验结果表明,在该算法控制下数据样本回溯期为10 min,低于传统算法,能够较好维护软件定义网络的运行安全性。

结合聚类边界采样的主动学习

主动学习是一种机器学习方法,需要选择最有价值的样本进行标注。目前,主动学习在应用时面临着一些挑战,其依赖分类器的先验假设,这容易导致分类器性能意外下降,同时需要一定规模的样本作为启动条件。聚类可以降低问题规模,是主动学习的一种有效手段。为此,结合密度聚类边界采样,开展主动学习方法的研究。针对容易产生分类错误的聚类边界区域,通过计算样本密度,提出一种密度峰值聚类边界点采样方法;在此基础上,给出密度熵的定义,并利用密度熵对聚类边界区域进行启发式搜索,提出一种基于聚类边界采样的主动学习方法。试验结果表明,与文献中的5种主动学习算法相比,该算法能够以更少标记量获得同等甚至更高的分类性能,是一种有效的主动学习算法;在标记不足,无标签样本总量20%的情况下,算法在Accuracy、F-score等指标上取得较好的结果。

基于模态分解与SRU网络的时间序列预测

时间序列预测在工业、农业、金融及军事等领域中具有重要的应用价值。为了进一步提高预测的可靠性和准确性,构建一种基于模态分解与SRU网络的杂交预测模型。首先,针对模态个数难以确定的问题,构建基于平均样本熵来确定模态个数的自适应变分模态分解(AVMD)模型,以减少不同频率上的混叠及降低随机噪声的干扰。通过在Adam算法中引入了随机调整参数,来提高SRU网络的训练速度及增强网络跳出局部最优解的能力。最后,发展一种基于AVMD与SRU网络的杂交模型。为评估提出的预测模型的可靠性和准确性,将之与一些最新预测方法做比较。电力负荷序列的实验结果表明,所提出的杂交预测模型具有较高的准确性和可靠性。

管制员个体工作负荷多维量化研究

为提高空管系统高效运行,聚焦管制员个体工作负荷建立量化模型;首先设计试验采集一线16名区域管制员的岗前与岗后各项指标数据,根据测试数据变化,选择出敏感变量,描述个体工作负荷;其次建立包含心理感知负荷、生理反应负荷与脑力工作负荷3个维度的综合评估指标体系,构建管制员个体工作负荷指数模型;然后通过熵权-客观组合法求解个体工作负荷指数最优权重,最终得出管制员个体工作负荷量化模型;最后进一步根据管制员个体工作负荷综合指数进行K-Means聚类分析,结果表明:管制员因个体不同岗后工作负荷存在差异。依据个体工作负荷指数大小,管制员可分为3类,A类管制员数量占总人数50%,岗后个体工作负荷增长最小;B类管制员数量占总人数43.75%,岗后负荷增长居中;C类管制员数量占总人数6.25%,岗后负荷增长最大,与教员对管制员能力的评分结果一致。

一种灰色关联分析优化ICEEMDAN的VP倾斜仪信号降噪模型

VP倾斜仪固体潮信号受仪器监测复杂环境限制,多含有大量环境噪声。为获得真实固体潮曲线,提出一种基于灰色关联分析优化改进的自适应噪声完备集合经验模态分解(ICEEMDAN)VP倾斜仪信号降噪模型(GRA-ICEEMDAN)。该方法首先将含噪信号进行ICCEMDAN处理,得到若干个固有模态函数(IMF),并依次排列与标记;然后基于这些IMF分别计算相关系数、互信息、R^(2)、Adj-R^(2)、MSE、SSE、RMSE、MAE、MAPE、样本熵等10个评价指标值,构建IMF可信度评价指标矩阵;最后借助灰色关联分析(GRA)计算各评价指标与不同IMF之间的关联系数和关联度,依据关联度大小对各个IMF进行排序,将排名靠前的IMF进行线性重构,即可完成信号降噪。仿真去噪实验和实测去噪实验均表明,GRA-ICEEMDAN模型优于卡尔曼滤波、70阶低通FIR滤波、Savitzky-Golay等经典降噪模型,能显著区分噪声成分和有效成分,原始信号分解后的重构误差与信号损失极小,可推广至其他仪器的复杂信号降噪中。

基于复杂纹理特征融合的材料图像分割方法

为解决材料图像分割中存在小样本、纹理复杂和数据分布不平衡的问题,抓住材料图像同相像素具有高度相似性的特性,提出一种基于复杂纹理特征融合的材料图像分割方法。在编码阶段,使用全卷积神经网络(FCN)作为基础网络,VGG16作为骨干网络;将改进的FCN的每层的特征图放入设计的级联的特征融合模块(CFF block),融合高低层语义信息;将融合的特征图放入多尺度学习模块(multi-scale block)进一步提取纹理特征。在解码阶段,对特征图施加注意力机制(Attention block),保留关键的特征图;针对材料图像中数据不平衡问题,采用并改进Dice损失,优化分割结果。通过对比实验和消融实验验证该方法的mIoU在多个数据集上均优于经典的深度学习方法。

基于模态分解及GRU-XGBoost短期电力负荷预测

精确的短期电力负荷预测能有效提高电力系统运营水平。针对电力负荷数据受多种因素影响,波动性和随机性强等问题,提出了一种基于模态分解及混合模型的负荷预测方法。首先,采用主成分分析法(principal component analysis,PCA)对负荷特征向量进行处理,去掉冗余信息,再用完全自适应噪声集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)将历史负荷分解为简化的几个子序列;其次,选择引入样本熵(sample entropy,SE)来计算子序列熵值,将相近的子序列重构得到随机、细节、低频和趋势分量后选用不同结构门控循环单元(gate recurrent unit,GRU)对不同分量类型进行预测,再使用极致梯度提升模型(extreme gradient boosting,XGBoost)对各分量残差进行拟合,各重组序列的预测值为GRU预测值与XBGoost拟合值之和,重组各序列得到最终预测值。选取3年时电力负荷数据进行实验,结果表明,所提模型的均方根误差(root mean square error,RMSE)、平均绝对百分比误差(mean absolutepercentage error,MAPE)和平均绝对误差(mean absolute error,MAE)分别为370.676 MW、99.07%和246.89 MW,与单一模型和混合模型相比,实现了评价指标的明显减少。

CEEMDAN-SE-WT降噪方法在航空发动机燃油流量信号中的应用

燃油流量信号是反映发动机状态和计算飞机排放物排放量的重要信号,但飞机飞行过程中传感器采集信号时不可避免地会受到外界环境以及内部因素干扰。提出一种结合样本熵(sample entropy,SE)的完全自适应噪声集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)与小波变换(wavelet transform,WT)的联合降噪方法。首先使用CEEMDAN对燃油流量信号进行分解得到本征模态分量,利用样本熵筛选含噪分量,并用相关系数与方差贡献率进行复核。对于含噪分量使用小波阈值降噪进行处理。最后将未处理的模态分量和完成降噪的模态分量重构得到最终燃油流量信号。通过与其他方法比较,CEEMDAN-SE-WT方法拥有最高信噪比为85.287,降噪后燃油消耗总量与飞机总重变化最为接近,可以认为该方法较大程度保留了燃油流量信号中的有效特征,为后续计算民机排放物排放总量提供了良好的数据支持。

基于VMD-SE的电力负荷分量的多特征短期预测

为提高电力负荷的预测精度,提出一种基于VMD-SE的电力负荷分量的多特征短期预测方法。首先采用变分模态分解(VMD)将原始负荷分解为一系列模态分量与残差,VMD的分解层数由样本熵值(sample entropy,SE)确定;然后对比原始负荷与模态分量的SE值,重构为平稳分量和波动分量,来降低运算规模;同时利用皮尔逊相关系数来筛选特征变量,删除特征冗余,建立灰狼算法优化后的支持向量回归模型(GWO-SVR)和长短期记忆神经网络(LSTM)分别对平稳分量和波动分量预测;最后以某地区2018—2020年用电负荷为例进行实验。实验证明:此模型精准度高达94.7%,平均绝对百分误差降低到2.98%,具有更好的精准性和适用性。

脑电信号多特征融合与卷积神经网络算法研究

针对脑电信号(electroencephalogram,EEG)运动想象中单一特征无法多维表征信号中的信息导致的分类准确率不高的问题,提出一种基于样本熵和共空间模式特征融合的特征提取算法。算法先对原始脑电信号进行小波包分解,从中选择包含μ和β节律的分量进行重构,然后分别提取重构信号的样本熵和CSP(common spatial pattern,CSP)特征,将两者融合组成新的特征向量,使用所设计的一维卷积神经网络对其进行识别获得分类结果。所提方法在2003年BCI Dataset III中获得了91.66%的分类准确率,在2008年BCI Dataset A中获得了85.29%的平均分类准确率。与近年来文献中提出的多特征融合算法相比,准确率提高了7.96个百分点。