在奇异值分解(singular value decomposition,SVD)中提出了一种矩阵递推构造和分解算法,利用SVD实现了一种类似于小波包的信号分解方式,称之为多分辨SVD包.推导了多分辨SVD包的分解和重构算法,并提出一种用二维数组来存储这种包的三维...在奇异值分解(singular value decomposition,SVD)中提出了一种矩阵递推构造和分解算法,利用SVD实现了一种类似于小波包的信号分解方式,称之为多分辨SVD包.推导了多分辨SVD包的分解和重构算法,并提出一种用二维数组来存储这种包的三维数据的方法,避免了对内存的浪费.实例结果表明,这种包对信号的微弱变化具有优良的检测能力,其检测结果无幅值和相位失真,并能精确定位微弱变化的位置,这种包也能有效提取复杂信号中的弱故障特征,在这两方面均明显优于小波包的处理结果.展开更多
为对变压器有载分接开关机械故障进行诊断,提出一种结合奇异值分解SVD(singular value decompo-sition)消噪与小波包WP(wavelet packet)消噪的信号特征提取方法。首先对信号进行小波包消噪,然后进行SVD二次消噪,将消噪信号进行经验模态...为对变压器有载分接开关机械故障进行诊断,提出一种结合奇异值分解SVD(singular value decompo-sition)消噪与小波包WP(wavelet packet)消噪的信号特征提取方法。首先对信号进行小波包消噪,然后进行SVD二次消噪,将消噪信号进行经验模态分解EMD(empirical mode decomposition),对得出的各阶固有模态分量进行希尔伯特-黄变换HHT(Hilbert-Huang transform)。数值仿真表明基于WP_SVD降噪的信号特征提取比小波包或SVD单独降噪的信号特征提取方法有效,并成功地将该方法应用到分接开关实际振动信号分析中。展开更多
针对电能质量的扰动检测问题,以电流信号为研究对象,提出结合形态滤波与多分辨率奇异值分解(singular value decomposition,SVD)包的电能质量扰动检测算法。根据形态学滤波器计算特点,采用余弦结构元素,对滤除噪声后的信号构造矩阵进行...针对电能质量的扰动检测问题,以电流信号为研究对象,提出结合形态滤波与多分辨率奇异值分解(singular value decomposition,SVD)包的电能质量扰动检测算法。根据形态学滤波器计算特点,采用余弦结构元素,对滤除噪声后的信号构造矩阵进行多分辨率SVD包分解,通过分解后的高频分量特征检测扰动,结合自适应阈值判断是否发生扰动,利用仿真对其进行验证。仿真实验结果表明:该算法相较于普通形态学与SVD方法有更好的抗噪能力,且可实现对扰动信号的快速、准确定位。展开更多
An analysis of the received signal of array antennas shows that the received signal has multi-resolution characteristics, and hence the wavelet packet theory can be used to detect the signal. By emplying wavelet packe...An analysis of the received signal of array antennas shows that the received signal has multi-resolution characteristics, and hence the wavelet packet theory can be used to detect the signal. By emplying wavelet packet theory to adaptive beamforming, a wavelet packet transform-based adaptive beamforming algorithm (WP-ABF) is proposed . This WP-ABF algorithm uses wavelet packet transform as the preprocessing, and the wavelet packet transformed signal uses least mean square algorithm to implement the ~adaptive beamforming. White noise can be wiped off under wavelet packet transform according to the different characteristics of signal and white under the wavelet packet transform. Theoretical analysis and simulations demonstrate that the proposed WP-ABF algorithm converges faster than the conventional adaptive beamforming algorithm and the wavelet transform-based beamforming algorithm. Simulation results also reveal that the convergence of the algorithm relates closely to the wavelet base and series; that is, the algorithm convergence gets better with the increasing of series, and for the same series of wavelet base the convergence gets better with the increasing of regularity.展开更多
文摘在奇异值分解(singular value decomposition,SVD)中提出了一种矩阵递推构造和分解算法,利用SVD实现了一种类似于小波包的信号分解方式,称之为多分辨SVD包.推导了多分辨SVD包的分解和重构算法,并提出一种用二维数组来存储这种包的三维数据的方法,避免了对内存的浪费.实例结果表明,这种包对信号的微弱变化具有优良的检测能力,其检测结果无幅值和相位失真,并能精确定位微弱变化的位置,这种包也能有效提取复杂信号中的弱故障特征,在这两方面均明显优于小波包的处理结果.
文摘针对电能质量的扰动检测问题,以电流信号为研究对象,提出结合形态滤波与多分辨率奇异值分解(singular value decomposition,SVD)包的电能质量扰动检测算法。根据形态学滤波器计算特点,采用余弦结构元素,对滤除噪声后的信号构造矩阵进行多分辨率SVD包分解,通过分解后的高频分量特征检测扰动,结合自适应阈值判断是否发生扰动,利用仿真对其进行验证。仿真实验结果表明:该算法相较于普通形态学与SVD方法有更好的抗噪能力,且可实现对扰动信号的快速、准确定位。
文摘An analysis of the received signal of array antennas shows that the received signal has multi-resolution characteristics, and hence the wavelet packet theory can be used to detect the signal. By emplying wavelet packet theory to adaptive beamforming, a wavelet packet transform-based adaptive beamforming algorithm (WP-ABF) is proposed . This WP-ABF algorithm uses wavelet packet transform as the preprocessing, and the wavelet packet transformed signal uses least mean square algorithm to implement the ~adaptive beamforming. White noise can be wiped off under wavelet packet transform according to the different characteristics of signal and white under the wavelet packet transform. Theoretical analysis and simulations demonstrate that the proposed WP-ABF algorithm converges faster than the conventional adaptive beamforming algorithm and the wavelet transform-based beamforming algorithm. Simulation results also reveal that the convergence of the algorithm relates closely to the wavelet base and series; that is, the algorithm convergence gets better with the increasing of series, and for the same series of wavelet base the convergence gets better with the increasing of regularity.