采用SVD-NMF的管道泄漏信号去噪算法

在管道泄漏检测中,压力信号中的噪声干扰会降低传统互相关法的定位精度。传统的去噪算法对环境的适应性差,去噪效果不理想。为此,提出了一种奇异值分解SVD( Singular Value Decomposition)与非负矩阵分解NMF( Nonnegative Matrix Factorization)相结合的管道泄漏信号去噪算法。该方法首先通过奇异值分解确定非负矩阵分解的阶数并对其初始化；然后,采用改进的非负矩阵分解算法对原信号进行迭代分解,获得去噪信号；最后,对去噪信号进行处理后通过互相关计算时延,并结合泄漏信号的传播速度实现泄漏定位。大量实验结果表明,SVD￣NMF算法能够显著降低迭代次数,提高去噪速度；同时在泄漏检测中,能够达到去除噪声干扰,提高定位精度的目的。

SVD-NMF based denoising algorithm for pipeline leak signal

In the pipeline leak detection,noise in pressure signals decrease the location accuracy of traditional cross￣correlation method. The traditional denoising algorithm cannot adapt to environment, and the effect is not good. Therefore,a denoising algorithm for pipeline leakage signal based on nonnegative matrix factorization ( NMF ) combined with singular value decomposition( SVD) is proposed. The leak signals are decomposed by singular value decomposition to determine the NMF’s order and initialize the matrix. The improved NMF algorithm is adopted to factorize the original signals and denoised signals are obtained. The time delay is calculated by cross￣correlation method,and the leak location is accomplished with the combination of the stress wave velocity. Experimental results show that the SVD￣NMF algorithm can reduce the number of iterations and enhance the speed of denoising signifi￣cantly. In the application of leak detection,the algorithm can remove noise and improve the location accuracy.