Research on Leakage State Classification of Pipelines Based on Wavelet Packet Analysis and Support Vector Machines

Aimed at the problem during the pipeline leakage state detecting process, that the datum which the sensor directly surveys are quite big and the characteristic are not strong, this paper brings forward one pipeline leakage state classification method, which combines the wavelet packet analysis and support vector machines. Through using the wavelet packet to the original data, carrying on the frequency band decomposing and the energy analysis, obtains the characteristic that can most reflect the classified essence. State sorter constituted by the support vector machines, only needs a few training samples, can make signal frequency band energy as the eigenvector to recognize and classify. The experiment datum shows that, this method effectively realizes the classified recognition of leakage state.