Font Size: a A A

Study On Fault Feature Extraction Of Rotating Machine Method Based On Local Mean Decomposition

Posted on:2014-02-17Degree:MasterType:Thesis
Country:ChinaCandidate:S S LiFull Text:PDF
GTID:2252330392464124Subject:Measuring and Testing Technology and Instruments
Abstract/Summary:PDF Full Text Request
Fault diagnosis is a science to identify the running state of device by studying thevarying information of the device running state. With the development of modernindustrial production, there are more and more attention paid to machinery equipmentfault diagnosis technology. The bottleneck of mechanical fault diagnosis is to exact faultfeatures from the vibration signal. The signal processing is the usual method for featureextraction. In nature, almost every vibration signal is non-stationary. How to analyze andto deal with those signals is undoubtedly vital for the success of fault feature extraction.In recent years, the time-frequency analysis method has been developing rapidly inthe field of mechanical fault diagnosis. The traditional signal analysis method based onFourier transform can’t process non-stationary signal perfectly. The paper focuses on thelocal mean decomposition (LMD) and the application in mechanical fault featureextraction.Firstly, the definitions of instantaneous frequency, simple component and multi-component signals, and modulation signal are discussed. The basic principles of LMD areanalyzed and the LMD algorithm is studied. Then it is compared with Empirical ModeDecomposition (EMD). The methods acquiring instantaneous frequency are discussed,such as the direct method, Hilbert transform based on LMD, and energy operatordemodulation approach. Due to their weaknesses, an energy operator demodulationalgorithm of three-point symmetrical differencing is proposed in this paper. Thesimulation shows that the method is efficient to reduce LMD end effect.Secondly, the influence of noise to LMD is analyzed. To eliminate phenomenon ofmode fission, a de-nosing technology, which is used to extract mechanical fault features,is discussed when semi-soft wavelet threshold is combined with LMD relevancy method.Theoretical analyses combined with mechanical fault signal are to verify the effectivenessof the method above in improving the performance of the time-frequency analysis.Lastly, based on the concept of time-frequency entropy, the local time-frequencyentropy is introduced, which is bonding with LMD to extract fault features. The results ofsimulation are presented to verify the theory analysis.
Keywords/Search Tags:Local Mean Decomposition, Fault feature extraction, Rotating machine, Endeffect, Mode mixing, Local time-frequency entropy
PDF Full Text Request
Related items