Wavelet Analysis for Diagnosing the Leaks of Pipeline

• Main Authors: Tseng Min-Hung, 曾明鴻
• Other Authors: 郭興家
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/10436669881227560502

碩士 === 國立成功大學 === 造船及船舶機械工程學系 === 87 === The thesis is to simulate actual leaks of pipeline from round hole leaks of pipeline systems in experiment. Measure the acoustic emission signal from the leak and use wavelet to analyze the signal. Use wavelet packet to decompose several frequency range of original acoustic emission signal in order to acquire the feature of acoustic emission signal, from which to proceed the training of back-propagation neural network(BPN). In this way, we can form the BPN to classify the different sizes of holes. The second part, we use two acoustic emission sensors to receive the same hole of acoustic emission signal, take wavelet to delete noises to reform the acoustic emission signal, and then proceed the cross-correlation function to receive the time differences of signals in order to judge the location of leaks. The result appears: extracting feature of pipeline leaks signal, wavelet packet is the more effective way of leaks analysis than fourier transform.