A pipeline leak detection method based on wavelet packet and BP neural network

According to the problems of pipeline leak detection existed in oil transport, a testing and positioning method based on wavelet packet and BP neural network is proposed. A Db4 wavelet packet is used to filter the noises of pressure signals, and extract its energy signals as the input vectors to train the BP neural network, which can be used to identify the pipeline´s working state and accurately locate the leak points of the pipe. The simulation experiment with water pipe shows that this method is accurate, reliable, and has strong adaptability.