Research Of Quantification Feature Extraction Method For Rotor Fault Signal Based On Entropy And EMD

With the development of transducer technology, measuring and testingtechniques, signal analysis, fault diagnostic techniques have more advance, itinclude four research contents: acquisition of signal, signal analysis and processing,fault diagnostic, information integration. There into, research of signal analysis andprocessing is the technology of feature factor extraction during the process of faultdiagnostic to obtain large relevance’s sensitive feature with system status. Theveracity of feature extraction can immediate impact the result of fault diagnostic.Aiming at traditional signal analysis method difficult to describe non-stationary ofrotor vibration signal and difficult to evaluate signal feature quantitatively, thisresearch put forward a method of quantitative feature extraction in the rotor faultsignal based on entropy and Empirical Mode Decomposition (EMD).Aiming at collected date from rotor experiment table, concrete researchescarried out and main results obtained are as follows:1) On the basis of bearing-rotor system model, research rotating machinery’sfault type and theory. Aiming at the characteristic of rotor fault vibration signal usemedian filter and wavelet filter pretreatment, provide original data for featureextraction. Research of this text signal analyse way—Empirical ModeDecomposition (EMD), analyze its essence and characteristic. Aiming at someproblems in the progress of EMD, put forward to a method what combined with“energy algorithm” and “same signal move algorithm”. Simulation signal indicatethat this method can solve iterations and the endpoint effect problems exactly andeffectively.2) Analysis research status of entropy in fault diagnosis. Study the nature ofentropy in-depth. Does a systematic research and analysis on the time domainsingular spectrum entropy, frequency domain power spectrum entropy,time-frequency domain wavelet energy spectrum entropy and wavelet space stateentropy. Programming the entropy program what based on LabVIEW.3) Put forward rotor fault signal quantization feature extraction methodcombined with EMD and entropy. Put forward an algorithm which combined “energyalgorithm” with “correlation coefficient algorithm” to determine IMFs which containthe key fault information. The experiments show that this method can extract and the school Foundation of Lanzhou University of Technology.characteristic data for follow-up study accurately. Calculate of four entropy whichcontains the main fault., time domain singular spectrum entropy, frequency domainpower spectrum entropy, time-frequency domain wavelet energy spectrum entropyand wavelet space state entropy, The result of this experiment show that this methodis used to extract quantitative feature in rotor system fault have good consequent,and own the function that can distinguish the difference from typical fault signalobviously.4) Based on the idea of information integration, selecting the IMFs whichcontains main fault information after EMD, calculate its singular spectrum entropy,power spectrum entropy, wavelet energy spectrum entropy to establishment of theentropy status spatial distribution model for the fault signal. Analysis shows that thisstatus spatial distribution model can realized of rotor fault pattern recognitionintuitively and accurately.5) Established a test system aim at typical rotor fault signal what base onLabVIEW. This system filtering original signal and analyze frequency and axistrajectory.In this text the fault signal rotor quantitative feature extraction for the purpose,aim at digital signal processing、information theory and intelligent diagnosis theory,research work deserves further study.