Research On Variable-scale Demodulation And Feature Extraction Algorithm For Gearbox With Gear Fault

Gearboxes are the key basic components in mechanical equipment,which have been widely used in many important areas to achieve power transmission and rotational speed changes,such as aerospace,shipbuilding,wind power,automotive,weaponry,construction machinery and precision machine tools.With the development of mechanical transmission equipment towards to high power density and large-scale,the working environment of gearbox has become more complicated and tough.The failure probability of gearbox increased,which has seriously affected the precision and reliability of mechanical transmission equipment.If the early fault vibration characteristics of gearboxes can be found in time,effective maintenance can be done to reduce the economic loss and catastrophic accidents.Therefore,an accurate extraction and identification of the early fault vibration characteristics of gearbox is one of the key factors for fault diagnosis success rate,and which has important theoretical significance and practical engineering value for improving the safe service performance of gearbox.Due to the influence of noise interfences,non-stationary operating conditions,coupled vibration sources,transmission paths,and interfaces of mechanical equipment,the signal-to-noise ratio,the time/frequency domain distribution and the detailed waveform of the periodic weak shocks induced by early fault have changed.The modulation scale of the fault vibration characteristics varies in real time with the operating conditions and other factors.Therefore,it is urgent to develop new variable-scale demodulation and extraction methods for the early fault vibration characteristics of gearboxes.Aiming to solve the problem of variable-scale demodulation and extraction methods for the fault vibration characteristics of gearboxes,a series of studies have been carried out,which includes the improvement of signal-to-noise ratio,variable-scale demodulation,feature extraction and and identification of the fault vibration characteristics.The main research contents includes:(1)For the elimination of operational noise of machinery,a new gobal optimal filter searching rule is proposed considering the time-varying statistical characteristics of noise,non-uniform frequency domain distribution,and compositional complexity.The basic idea of the searching rule is “discover the global peaks,find the local optimals” and “survival of the fittest”.Based on this idea,a new dynamic-iteration evolutionary digital filter noise cancellation model is presented.The new model dynamically adjusts filter population evolution rules in the searching process of the optimal filter,and realizes the transition from global cross-peak search to local refinement search on the noise reduction performance surface.The new model overcomes the convergence speed and premature problems of the traditional evolutionary digital filter method.The noise calcellation performance and convergence efficiency of the proposed method have been validated by both simulated signal and experimatal signal.(2)For the optimal demodulation subband selection,a new subband evaluation index in envelope spectrum is proposed to qualitify the fault-related vibration component in each subband considering the limitation of spectral kurtosis subband selection index.Therefore,a variable-scale searching model for the fault characteristics demodulation subband is established based on genetic algorithm.The new model overcomes the problems of fast kurtogram method of poor correlation to fault vibration and easily affected by strong noise,the effectivenss of the new model are validated by both simulated signal and experimental signal in planetary gearbox.(3)For the fault characteristics demodulation in case of variable working conditions,traditional fixed-scale demodulation methods are usually degraded to get used to the variable rotational speed.A new multi-scale time-frequency impulse demodulation method is put forwards based on empirical mode decomposition and time-frequency analysis method,the impulse envelope signal is detected from time-frequency domain distribution after the sensitive instinct mode functions are selected and reconstructed.The new method settled the problem of fixed-scale demodulation method in adaptively changing frequency scale in case of the variable rotational speed,and which has been verified by some experiments.(4)For the fault sensitive feature selection,traditional distance evaluation technique tends to select redundant features because it does not consider the relationships between features.Aiming to overcome this problem,an enhanced distance evaluation technique is presented to select the sensitive feature subset from the whole extrated feature set,both irrelavent and redundant features are removed simultaneously,then a gear faults classification algorithm model is established based on k-means clustering method.The effectiveness of the proposed method is validated by the classification of different located faults and different faults degrees.(5)Based on the variable-scale demodulation and extraction methods for the early fault vibration characteristics of gearboxes,a romote on-line health monitoring system for gearbox is established,in which three kinds of modules are developed: signal acquisition,signal analysis and operation condition analysis.Both the signal acquisition,signal transmission,denoising,variable-scale demodulation analysis,feature extraction and condition identification are realized in the system.And,the application capability in partical engineering of the variable-scale demodulation and extraction method for the early fault vibration characteristics is validated.