Research On Pipeline Leak Signal Processing Method Based On Nonnegative Matrix Factorization

Pipeline leak detection technology is an important sector of pipeline transport.Improve the accuracy of leak positioning can reduce the work load of the on-site maintenance workers.Pipeline leak detection system based on negative pressure wave is widely used in pipeline leak detection because of its low cost,convenient maintenance and other advantages.However,negative pressure wave method has problem in high false alarm rate due to the irregularity and low signal-to-noise ratio of pipeline pressure signal,and results in low detection efficiency and precision.To solve this problem,the denoising algorithm based on nonnegative matrix decomposition is proposed.Based on that,leak detection and location method is studied.This paper will focus on the following works:Firstly,some pipeline leak detection method were mentioned,and the principle of negative pressure wave method of pipeline leak detection system and relative research focus were introduced.The impact of noise on negative pressure wave method was analyzed.The noise reduction requirement of negative pressure wave detection was put forward,and scholars’ research on the denoising algorithm was presented.Secondly,the non-stationary signal denoising method was discussed.Due to the traditional denoising algorithm d id not fit to the negative pressure wave signal,the denoising algorithm based on nonnegative matrix decomposition(NMF)was proposed.The basic algorithm and iterative formula of the nonnegative matrix decomposition were introduced,and according to the problem of poor effect when nonnegative matrix decomposition process complex data,nonnegative matrix factorization combined with singular value decomposition(SVD)is proposed in order to improve the nonnegative matrix decomposition algorithm’s iterative speed,and it’s called SVD-NMF algorithm.Experiment results show that the SVD-NMF algorithm can effectively denoise the non-stationary signal.Thirdly,in view of the traditional cross-correlation method cannot calculate the accurate time delay,a kind of time delay estimation method based on weighted residuals was proposed.This method processes the signal after denoised,the amplitude of each moment is related to the pressure all time before.The processed signal becomes sparser,and cross-correlation calculation can get more accurate time delay estimation.Experiment results prove that weighted residuals algorithm can improve the precision of time delay estimation.Finally,liquid pipeline experimental platform was set up in the laboratory.The signals of collection normal operation,large leakage,small leakage and operating disturbance were collected and analyzed respectively.Experiment results show that the proposed pipeline leakage detection and localization algorithm provides more accurate positioning.At the same time,the algorithm can distinguish between operating disturbance and leakage,reduce false alarm rate.