储罐底板声源辨识及定位方法研究

储罐底板腐蚀是多声源问题,即在不同位置的腐蚀源可能同时发射应力波。这些声源信号有时会重叠被传感器接收,从而影响定位的可靠性。为此本文基于平面声发射源能量定位方法的基本理论,进行了模拟储罐底板定位实验,提出了能量定位系数的修正方法。同时通过对实验数据分析,发现快速独立分量分析(FastICA)方法可以将同种声源混合信号进行有效分离,并且基本保持原有波形特征,相干系数法可以实现对分离后的同源信号进行聚类,进而应用改进能量定位方法对声发射源进行定位,从而对声源辨识,判断事件集中度提供依据。     

基于Khatri-Rao子空间和传播算子的宽带声源波达方向估计算法

本文提出基于Khatri-Rao子空间和传播算子的宽带声源波达方向估计算法。该算法将声源不同频率处的协方差矩阵变换重排为一个高维矩阵,然后利用传播算子方法估计宽带声源波达方向。该算法计算复杂度介于聚焦Khatri-Rao子空间和相干子空间算法之间。仿真和实验结果表明,该算法在降低计算量的同时,估计误差与聚焦Khatri-Rao子空间算法相近,远小于相干子空间算法。     

远近场混合源的波束解卷积定位

提出了定位远近场混合源的波束解卷积技术,针对非相干远近场混合声信号的线列阵观测结果,推导了其常规波束形成(CBF)空间谱中固有的广义二维卷积数学关系,利用Richardson-Lucy算法实现波束能量聚焦以获得近场目标的精确空域参数估计,通过混合源协方差矩阵向近场流形的正交补空间投影操作提取远场分量,并分析得到其内在的一维卷积关系,然后通过角度域波束解卷积进行远场信号的波达估计。仿真分析表明,所提方法提升了CBF谱的空域分辨力,通过投影映射隔离近场分量后实现了混合源的分离。与现有方案相比,所提算法针对远场信源可实现10 dB的背景噪声级抑制。     

基于矢量阵的运动声源柱面聚焦定位方法试验研究

本文基于被动合成孔径原理, 在建立运动声源矢量阵近场柱面聚焦测量模型的基础上, 分别研究了适用于单频线谱信号和宽带连续谱信号的矢量阵柱面聚焦定位方法, 通过数值仿真计算了该方法在多种误差条件下的定位精度, 并进一步通过舱段模型试验对该方法的工程实用性和正确性进行了详细的分析和论证. 舱段模型试验结果表明, 柱面聚焦定位结果与壳体振动能量分布规律符合较好, 该方法不仅能真实反映声源位置信息, 而且能反映不同频带内声源能量分布的相对大小, 具有良好的定位效果.     

浅海环境中的匹配模态空间检测器

将水下声传播规律融入到算法设计中可以有效提高被动声呐目标检测性能。当声源位置未知时,广义似然比检测器和贝叶斯检测器分别通过搜索和积分的方式来消除声源位置不确定性的影响。但是,基于有限个信号波前实现的广义似然比检测器和贝叶斯检测器在某些声源位置上存在性能大幅下降的问题。为此,利用水下声传播的物理特性,提出了一种稳健的子空间检测器——匹配模态空间检测器,稳健的意义在于:当阵列获取到的辐射声信号能量给定时,检测器可以在不同声源位置情况下提供相同的检测性能。该检测器通过模态空间一定程度上利用了海洋环境知识,获得了比具有相同稳健性的能量检测器更好的检测性能。典型浅海环境中的仿真实验对比结果表明:匹配模态空间检测器相比广义似然比检测器和贝叶斯检测器的峰值性能下降较小、所需的计算量更少、对环境失配的宽容性更好。      

基于传播算子的宽带源相干信号子空间测向方法

将窄带传播算子方法引入宽带源测向的相干信号子空间处理中。首先提出一种新的聚焦矩阵的计算方法;该方法不需要进行角度预估,也不需要任何矩阵分解,对信号方向不敏感,运算量小。在此基础上,根据相干信号子空间处理方法的原理,对聚焦的空间协方差矩阵作传播算子方法的处理。整个过程避免了矩阵分解,因而运算量降低,仿真结果证明了该方法的有效性。     

基于Group Lasso的多重信号分类声源 定位优化算法*

多重信号分类算法因其抑制噪声能力强、计算速度快等优点,在声源定位领域得到广泛应用。但该算法在中低频段分辨率及聚焦性能较差。针对该问题,提出一种基于Group Lasso的多重信号分类优化算法。该算法将多重信号分类算法输出值作为初始值,并在Group Lasso算法组间计算时对目标信号进行稀疏、在组内计算时对该组信号进行平滑及阈值截断。仿真结果表明:该优化算法在中低频段可明显提高多重信号分类算法分辨率,同时改善因扫描位置与声源面位置不重合引起的聚焦性能下降问题。     

阵元随机均匀分布球面阵列联合噪声源定位方法

基于球面传声器阵列的噪声源定位方法,设计加工了阵元随机均匀分布64元球面传声器阵列,研究了球面近场声全息和球谐函数模态展开聚焦波束形成联合噪声源定位识别方法,对算法的性能进行了仿真分析,并利用球面传声器阵列进行了噪声源的定位识别试验.研究表明,阵元随机均匀分布球面阵列具有全空间稳定的目标定位性能,球面近场声全息对低频近距离声源具有较高的定位精度,球谐函数模态展开聚焦波束形成对高频远距离声源具有较高的定位精度,将两种方法联合进行声源的定位识别,可以在较小孔径的球面阵列和较少阵元的条件下,在宽频带范围内获得对目标声源良好的定位性能.     

滚动轴承声信号特征提取和诊断试验研究

为解决振动检测方法不能有效识别低速旋转机械滚动轴承故障问题,利用声发射检测方法,建立了滚动轴承低速声发射信号采集试验装置,对模拟人工缺陷滚动轴承声发射信号进行了采集,进而对滚动轴承声发射信号进行总体平均经验模式分解,结合能量矩及相关系数法综合判断分解后各模态分量的真伪,据此提取出特征信号并做出其局部Hilbert边际谱,最后对滚动轴承各种故障模式进行诊断。试验结果表明该诊断方法能准确识别滚动轴承声发射信号故障频率,依据特征频率及幅值大小可对低速滚动轴承故障进行有效诊断。     

高精度球面阵聚焦声源定位方法研究

提出了一种高精度高空间分辨率球面阵聚焦声源定位方法——虚拟源法。该方法通过球面阵波束扫描获得实际声源的空间聚焦谱,并假定各扫描点为虚拟声源,将实际声源聚焦谱看作是全体虚拟源共同作用的结果,由此得到各虚拟源对声场的贡献量,从而可实现声源精确定位。仿真研究分析了频率,阵列孔径,声场模态阶数,信噪比等参数对声源定位性能的影响,并与常规算法进行对比。结果显示,该方法不受频率和阵列孔径的限制,避免了空间“混淆”,能够进行高精度高分辨率声源定位,并具有良好的背景噪声抑制能力。      

Low speed bearing fault diagnosis using acoustic emission sensors

In this paper, a new methodology for low speed bearing fault diagnosis is presented. This acoustic emission (AE) based technique starts with a heterodyne frequency reduction approach that samples AE signals at a rate comparable to vibration centered methodologies. Then, the sampled AE signal is time synchronously resampled to account for possible fluctuations in shaft speed and bearing slippage. The resampling approach is able to segment the AE signal according to shaft crossing times such that an even number of data points are available to compute a single spectral average which is used to extract features and evaluate numerous condition indicators (CIs) for bearing fault diagnosis. Unlike existing averaging based noise reduction approaches that require the computation of multiple averages for each bearing fault type, the presented approach computes only one average for all bearing fault types. The presented technique is validated using the AE signals of seeded fault steel bearings on a bearing test rig. The results in this paper have shown that the low sampled AE signals in combination with the presented approach can be utilized to effectively extract condition indicators to diagnose all four bearing fault types at multiple low shaft speeds below 10 Hz.     

Spatial time-frequency DOA estimation based on joint diagonalization using Jacobi rotation

Direction-of-arrival (DOA) estimation consists of locating closely spaced sources impinging from different directions in the presence of considerable noise or interference. Recently, this problem has been solved using spatial time-frequency distribution (STFD) information that is available in the array signals. In this work, a new joint diagonalization approach based on Jacobi rotation, which can efficiently combine all of the relevant STFD points, is proposed to achieve superior DOA resolutions and suppressed sidelobes in the spatial spectrum. It is discovered that our proposed Jacobi technique leads to a non-orthogonal joint diagonalization structure and can avoid pre-whitening the signal component of the observation. In addition, we further adopt the minimum variance distortionless response (MVDR) processor instead of the MUltiple SIgnal Classification (MUSIC) algorithm to avoid the estimation of the signal subspace and noise subspace in the time-frequency DOA domain. Finally, computer simulations of several frequently encountered types of challenging scenarios (such as low SNR and coherent arrivals) show that significant improvements are achieved by our proposed approach in comparison to the existing techniques.     

一种改进的稀疏近似最小方差DOA估计算法研究

针对稀疏信号的超分辨方位估计问题,提出一种可变因子的稀疏近似最小方差算法(α-Sparse Asymptotic Minimum Variance,简记为SAMV-α)。该算法利用一个折衷参数进行最大似然估计值和稀疏性能的折衷处理,在迭代过程中改变稀疏近似最小方差算法(Sparse Asymptotic Minimum Variance,SAMV)的指数因子,得到强稀疏性能和超低旁瓣的方位谱图,实现邻近目标的超分辨方位估计和相干处理性能,且无需预估角度和信源数目等先验信息,并且折衷参数的取值为0到1之间,取值区间明确,避免了稀疏信号处理算法中正则因子选取困难的弊端。计算机仿真表明SAMV-α算法方位估计性能明显优于波束扫描类算法和子空间类算法,与同类型稀疏信号处理类算法相比仍具有较高的方位估计精度,同时对于邻近声源分辨能力,SAMV-α算法较SAMV-1算法性能提高约3dB。海上试验数据处理给出了分辨率更高的方位时间历程(Bering-Time Recording,BTR)图,有效验证了SAMV-α算法的性能。      

宽带双曲调频信号参数化时频分析波达方向估计

针对有源探测或脉冲侦查中双曲调频信号的波达方向估计问题,提出了基于参数化时频变换(PTFT)的多重信号分类(MUSIC)测向算法,简称PTFT-MUSIC算法。该算法由发射信号确定针对双曲调频信号的参数化变换核,对接收信号进行频域参数化时频变换,利用获得的时频分布建立阵列信号时频分布模型,并以此模型设计基于时频分布矩阵的MUSIC算法以实现双曲调频信号的波达方向估计。通过仿真和实验对该算法的估计误差和多目标分辨性能进行了分析,仿真和海上实验结果表明:相比现有的时频MUSIC算法,PTFT-MUSIC算法能有效提高空间谱分辨率和波达方向估计性能,同时该算法拥有对特定调频信号筛选性,结合时频域滤波算法能有效抑制相干直达波干扰,应用于多基地声呐系统时有效提高了声呐定位性能。     

Leak detection in water-filled plastic pipes through the application of tuned wavelet transforms to Acoustic Emission signals

A new method to detect leakage in a water-filled plastic pipe is proposed. In this method, a leakage signal-signature in time domain is first captured by monitoring the Short Time Fourier Transforms (STFT) of AE (Acoustic Emission) signals over a relatively long time-interval. The captured signal is then used to find a mother wavelet (tuned wavelet) for the best signal localization in time and frequency domains. The technique for AE signal detection using tuned wavelet is then described. Practical application of the method proposed herein is then presented using a water-filled plastic pipe as a case study. Signals generated from this experimental setup are collected to identify leakage signal-signatures from other interfering signals (background, pipe natural frequency, splash and environmental noise). The results of the experiment prove that using tuned wavelet, AE events can be detected and identified precisely in time. In addition, sources of signals due to leakage and their respective energy levels can also be recognized.     

Identification of two simultaneous partial discharge sources in an oil-pressboard insulation system using acoustic emission techniques

Insulation failure is one of the major causes of catastrophic failure of transformers. It is established that partial discharge (PD) causes insulation degradation and premature failure of insulation. In power apparatus, more than one PD source may be active simultaneously. The nature of insulation degradation for multiple PD sources is different from that due to single PD source. Therefore, it will be helpful for severity assessment of insulation degradation, if the number of active PD sources are identified and classified. This paper presents a method for identification and classification of two simultaneously active PD sources using acoustic emission techniques. The acoustic emission (AE) signals are measured for laboratory simulated PD in an oil-pressboard insulation system for three different electrode systems. The measurements of partial discharge acoustic emission (PDAE) signals are carried out for single PD source and for two simultaneous PD sources. The measured signals are analyzed using discrete wavelet transform (DWT), box counting fractal dimension and lacunarity. Box counting fractal dimension and lacunarity are calculated for DWT decomposed signal of major frequency band. Energy distribution in different frequency bands of DWT decomposed signal along with box counting fractal dimension and lacunarity is used for classification of two simultaneous PD sources.     

基于信号子空间增强和端点检测的大地电磁噪声压制

为了进一步保留大地电磁低频段的有用信息、提高矿集区复杂噪声环境下大地电磁测深深部探测能力,在形态滤波的基础上结合信号子空间增强和端点检测做二次信噪分离处理.首先,针对形态滤波预提取的噪声轮廓运用信号子空间增强分离出信号子空间和噪声子空间.然后,将信号子空间和重构信号相结合并将噪声子空间置零.最后,借鉴端点检测做后处理,以识别波形突变的起止点.仿真结果表明,卡尼亚电阻率曲线在低频段的数据质量得到了明显改善、视电阻率值相对稳定;有效地补偿了形态滤波处理过程中损失的低频有用信号,其结果更加真实地反映了测点本身所固有的大地电磁深部构造信息.     

TERFDA: Tensor Embedding RF Domain Adaptation for varying noise interference

Radio frequency machine learning (RFML) can be loosely termed as a field that machine learning (ML) and deep learning (DL) techniques to applications related to wireless communications. However, traditional RFML basically assume that the data of training set and test set are independent and identically distributed and only a large number of labeled data can train a classification model which can effectively classify test set data. In other words, without enough training samples, it is impossible to learn an automatic modulation classifier that performs well in varying noise interference environment. Feature-based transfer learning minimizes the distribution difference between historical modulated signal data and new data by learning similarity-maximizing feature spaces. Therefore, in this paper, Dynamic Distribution Adaptation (DDA) is adopted to address the above challenges. We propose a Tensor Embedding RF Domain Adaptation (TERFDA) approach, which learns the latent subspace of the tensors formed by the time–frequency maps of the signals, so that use the multi-dimensional domain information of the signals to jointly learn the shared feature subspace of the source domain and the target domain, then perform DDA in the shared subspace. The experimental results show that under the modulated signal data, compared with the state-of-the-art DA algorithm, TERFDA has less requirements on the number of samples and categories, and has superior performance for confrontation the varying noise interference between source domain and target domain.     

基于多标签共享子空间学习和内核脊回归的空谱分类算法

针对高光谱图像维度高、地物间非线性可分造成的分类精度低等问题,提出一种基于多标签共享子空间和内核脊回归的空谱分类算法.该算法利用内核脊回归将地物相近像素在线性空间的不可分特征映射到高维空间中,实现分类特性在高维空间下的有效分离,以提高地物相近特性的区分精度;同时将高维样本数据映射到低维共享子空间中,在低维环境下以多类标为指导,引入低秩矩阵建立类别标签与共享空间的预测关系,挖掘多标签间的共同特性,提高融合利用多类别间的共同属性提高高光谱图像的分类精度;最后利用奇异值分解迭代法求解目标函数,一定程度上加速参数求解.在Indian Pines和Pavia University两组高光谱数据集上进行仿真实验,实验结果表明,与其他同类算法相比,在低样本比例下,本文算法在总体分类精度、平均分类精度和Kappa系数等评价指标上至少提高4.76%、4.24%和5.19%,与非内核化的算法相比,本文算法在基本不增加运行时间的情况下总体分类精度、平均分类精度和Kappa系数至少提高2.92%、2.8%和3.48%.