Research On The Method Of Planetary Gear Wear Fault Diagnosis Based On High-order Spectrum Theory

Planetary gearbox plays a very important role in ensuring the safe and efficient operation of mechanical equipment.In order to avoid unnecessary losses,it is often necessary to carry out real-time monitoring and regular maintenance in engineering practice.When the planetary gearbox fails,most of the measured signals have the characteristics of nonlinear,non-stationary and non-Gaussian distribution.It can effectively eliminate the Gaussian noise in the signals and can also extract the coupling components of the signals while reflecting the fault characteristics of the signals with higher-order statistics theory.In this paper,the research on wear fault diagnosis and identification of planetary gear is carried out with the three-stage planetary gearbox based on high order spectrum theory.It mainly involves the following four aspects of research work:(1)The theory of high order statistics is introduced.The definitions and differences are described between high-order moment and high-order cumulant in detail.The definition of high-order spectra is introduced.And the properties are analyzed and verified about bispectrum,and diagonal slice spectra by simulation signals.(2)Analysis and research on the structure of planetary gearbox and the denoising algorithm of vibration signal.The structure and characteristic frequency are introduced about planetary gearbox in experiment.And It is introduced the theory of wavelet denoising algorithm.The improved Bayes threshold denoising algorithm is applied to the noise reduction of gear vibration fault simulation signal and actual wear fault signal.The experimental results show that the improved Bayes threshold denoising algorithm improves the signal-to-noise ratio of the signal in this paper.(3)Analysis and research on the feature extraction method of planetary gear wear fault.The EMD theory and IMF component selection rule are introduced.The diagonal slice spectrum analysis is carried out about the selected reconstructed signal.The experimental results show that the fault frequency,meshing frequency and related modulation frequency of gear can be extracted effectively.The gear faults can be identified and it is better than using bispectrum,diagonal slice spectrum alone and traditional power spectrum algorithm by the EMD-diagonal slice spectrum algorithm.(4)The gear wear degree identification method based on diagonal slice spectrum-Elman neural network is proposed.The theory of Elman neural network,the setting of parameters and the requirements of input and output vectors are introduced in this paper.The amplitudes of feature frequencies of planetary gears with different wear degreeswere extracted as input feature vectors of Elman neural network.Then the input feature vectors are trained and recognized by Elman neural network.The result shows that the wear degree of planetary gear can be distinguished effectively by diagonal slice spectrumElman neural network method.