Acoustic Emission and Dual-Tree Complex Wavelet Transform with Soft Threshold De-Noising to Enhance Pipeline Leak Detection and Location

Abstract

Water leakage control is a subject of growing interest among researchers and pipeline owners. Various water leak detection methods have been developed and acoustic leak detection technique has manifested as a decent approach. However, the demerit of this method is that the collected signal always interferes with noises. This reduces the accuracy of leak detection and triggers a false alarm. Traditional noise reduction methods such as digital filtering and wavelet de-noising are not efficient to eliminate noise from interested signals. To overcome this drawback, an improved de-noising method derived from Dual-Tree Complex Wavelet Transform (DTCWT) and a soft threshold proposed in this paper to reduce the noise and consequently minimize false trip rate. The signal is decomposed by DTCWT to separate the raw signal into different frequency bandwidths. Then, soft threshold method is employed to suppress noise by rescaling the signals which are lower than the prescribed threshold to zero. Experimental results demonstrate that the proposed de-noising method performs well than ordinary wavelet de-noising. The accuracy of leak localisation can be increased by twice. The method accentuates the cross-correlation peak related to the time delay for leak localization and thereupon increases the accuracy of leak locating. The proposed method outperforms other methods in term of detectable range of leak source as it can detect leak up to 24 m whereas the conventional way only can detect up to 14.5 m.