Abstract: The low signal-to-noise ratio makes the signal processing of the leak detection of pipeline difficulty. The adaptive noise cancellation based on neural network is proposed to achieve the H2 optimal reconstruction and a desired robust against the effect of uncertainties in signal processing. In order to implement leak detection and location of pipeline correctly, multi-sensor signals are employed to overcome the shortcomings of creditability in case of single-sensor signal. The designed method can adjust the weights of linear combiner (LC) adaptively, and overcome the shortcomings of the method based on the averaging method of all sensor signals, which may lose accurate information and cause a large estimated error. The design method can not only obtain simplicity of implementation and conservation of operation time but also achieve better reconstruction performance than the method in the pressure and flow signal processing of the leak detection of pipeline.