Leak detection method for the jet fuel pipeline based on IUPEMD and DTWSVM

Jet fuel pipeline leakage will cause environmental pollution and safety-related accidents; therefore, the leak detection of jet fuel pipeline is a crucial for pipeline management. Compared with negative pressure waves, acoustic waves exhibit better attenuation resistance and longer propagation distance. However, acoustic waves are easily disturbed by noise, causing the acoustic signals to mix with a large amount of noise and reducing the detection system's accuracy to identify pipeline leaks. An improved uniform phase empirical mode decomposition (IUPEMD) denoising method is proposed in this paper. Compared with other denoising methods, intrinsic modal functions with more leakage information can be selected according to the similarity coefficient for signal reconstruction. The reconstructed signal retains the leak information to a greater extent, making the noise content extremely low, which can effectively improve the leak identification accuracy of the leak detection system. To accurately determine the leakage of pipeline and solve the problem of low accuracy of recognition model, this paper establishes a deep learning twin support vector machine (DTWSVM) for identifying the state of pipeline based on deep learning and twin support vector machine, which can automatically extract the leakage feature information and accurately determine the leakage of pipeline based on the feature information. The experimental analysis demonstrates that the IUPEMD denoising method can effectively filter the noise in the signal. The DTWSVM model showed very high recognition accuracy, and its leakage recognition accuracy can reach 99.6%.