Research On Damage Feature Extraction And Application Of Pipeline Based On HHT

Recent years, with the global economic rapid development, pipeline system haswidely applied from the daily life, industrial equipment, engineering machinery,transportation to the national defense fields. Pipeline system is a necessary part toensure equipment work normally. Pipeline is known as the "blood" or "lifeline" ofindustrial equipment. It is an important part of many field such as aerospace,transportation, engineering machinery. Thus it can be seen that pipeline acts animportant role in various fields. So it will directly affect various fields of productionand living and cause huge economic losses even serious accident whether the pipelinesystem is work normally.Hilbert-Huang Transform(HHT) is an effective signal analysis method which issuitable for non-stationary signal. This paper uses the HHT to extract the damagefeature of pipeline through analyzing the pipeline system’s signal, and then providetechnological support to damage detection of aero-engine pipeline.The main work in this paper is as follow.(1)We studied several common signal analysis methods such as Short TimeFourier Transform, Wavelet Transform and so on. Then we pointed out theshortcomings of these analysis methods when used to analyze non-static signal. Thispaper mainly studied the principle of Hilbert-Huang Transform and its superiority inanalysis of non-static signal. And it analyzed some problem in the HHT such asendpoint effect, mode mixing effect and stoppage criterion. We mainly studied themethod to eliminate the mode mixing effect and endpoint effect. We use theimproved method--EEMD and achieve good result.(2)Aimed at damage detection, we studied the smoothing character of EMD andused it to denoise the signal. The paper studied using EMD to decomposite signal toget the intrinsic mode function(IMF). Obtain The Hilbert spectrum by HHT andanalyze the Hilbert spectrum of signal. By comparing the normal status and thedifferent damage condition get damage feature vector. Analysis method include theHHT spectrum, marginal spectrum and time-frequency entropy. Through thesimulation and experimental data, comparison and analysis the effect of differentfeature extraction methods. (3)We analyzed aero-engine pipeline damage type and the possible cause of thepipeline damage. We collected the signal about vibration from the built aircraftengine experimental platform. Then we analyzed the signal of pipeline in the normalstate and damage condition when aero-engine worked in different speed. The signalanalysis method is HHT. We get the damage feature extraction by analyzing thesignal in different condition. Finally, we studied the application of the damage featureextraction. Combined with the neural network for damage identification, thecharacteristics of different damage are trained by neural network, so as to identify thedifferent damage.