Research And Application Of Non-stationary Signal Based On HHT Theory

In 1998,the American Academy of Engineering academician Norden E.Huang proposed a new signal processing method,Hilbert-Huang Transformation(HHT).The method mainly includes two parts: empirical mode decomposition and Hilbert spectrum analysis.The core content is the natural mode function and instantaneous frequency of the signal obtained by empirical mode decomposition.This method has been widely used in geology,medical,engineering and other fields.The basic theoretical concept and algorithm flow of HHT method are introduced in this paper.The experimental results of decomposing linear/nonlinear,stationary/nonstationary signals by HHT algorithm show that the end point value cannot be obtained because the three spline interpolation takes the signal envelope,resulting in the endpoint.Envelope information processing is inaccurate,resulting in end-effect effects at both ends of the signal and overshoot and undershoot in the signal.Two improved HHT methods are proposed in this paper,namely the HHT method based on BP neural network and the HHT method based on multi-point mirror extension.In the improved HHT method based on BP neural network,the BP neural network is used to predict the new extreme point of the signal to be decomposed,and the extreme points required for the three spline interpolation are intercepted.The HHT method is improved in multi-point mirror extension.In the direct use of the multi-point mirroring extension value as the extreme point required for three spline interpolation,the empirical mode decomposition is performed after the new extreme point is acquired to obtain the intrinsic mode function.Through the above two improved methods,the endpoint effect The problem has been significantly improved.At the same time,in order to verify the false components generated in the signal decomposition process,the correlation coefficient method and the normalized mutual information theory method are used to judge the intrinsic mode function.The experimental results show that the normalized mutual information is utilized.The theory can clearly identify the false signals in the improved HHT algorithm and verify the improved performance of the algorithm.Marine diesel engine has non-stationary and non-linear signal characteristics,and the vibration signal of marine diesel engine cylinder head is studied.The improved HHT method and HHT method are compared and analyzed from the two aspects of the intrinsic mode function and the instantaneous frequency.The simulation results show that the improved HHT method can recover the original signal well and effectively solve the end effect problem.