A practical preprocessing treatment for pipeline leak locating improving

Real leak signals, acquired in industrial pipeline systems may manifest two typical non stationary signatures, due to the disturbing noise: abrupt amplitude random changes on one side and a time varying mean on the other side. This paper proposes a pre-processing algorithm for extracting stationary information from the received signals, in order to improve the leak location on the pipe. This method combines the wavelet decomposition technique, in order to remove the random varying mean, with a segmentation algorithm, based on computing the stationarity index (SI), for avoiding the amplitude bursts. Comparative results show that, in addition, by using a pre-whitening filter and closing the pipelinepsilas end, improved estimates can be obtained.