Non-intrusive acoustic detection method for pipeline leak of hazardous substances

For pipeline leak detection, the intrusive method will not only destroy the integrity of a pipeline but also cause accidents due to hidden dangers such as weld quality. Therefore, in this paper, a non-intrusive acoustic leak detection method based on wireless sensor networks is proposed. Firstly, a non-intrusive acoustic detection device is designed, which is easy to install and adjust. Secondly, the simultaneous signal sampling of each wireless sensor node is realized by the time synchronization method combining software and hardware. Finally, based on support vector data description (SVDD), a pipeline leak diagnosis model is established with the input feature vector consisting of peak coefficients, kurtosis and wavelet packet energy fraction of acoustic signals. According to the test result, the proposed method has high sensitivity, accurate leak diagnosis, precise location and low false/missing alarm rate. It provides a novel and effective technical means for the pipeline leak detection of hazardous substances.