Research On The Detection And Extraction Of Weak Mechanical Impact Signal

Mechanical equipment will produce vibration and shock during operation.These vibration signals contain a great deal of equipment status information.The analysis and processing of these vibration signals are of great significance to the condition monitoring and fault diagnosis of mechanical equipment.However,in practice,due to the complex environment,the measured vibration signal often contains a lot of noise,making the characteristic signal containing equipment state information relatively weak or even submerged,which greatly affects the monitoring and fault diagnosis of machinery and equipment.In response to this problem,this paper carried out a PhD thesis named 'Research on the detection and extraction of weak mechanical impact signal' supported by National Natural Science Foundation of China(No.51175466)named Extraction and Recognition of Weak Impact Singal.The weak impulse signal detection and extraction algorithm under strong background noise is systematically studied.The weak impulse signal analysis and processing methods are proposed respectively from the detection of weak signals,the separation of multi-source fault signals,the extraction of time-domain and frequency-domain features.The main contents of the thesis are as follows.Aiming at the problem of weak impact signal detection and extraction,a weak impact signal detection method based on adaptive Morlet wavelet transform is proposed.The proposed method optimizes the central frequency and bandwidth parameters of the mother wavelet based on the minimum Shannon entropy criterion.Several wavelet coefficients are obtained by wavelet transform of impulsive signal with strong noises using Morlet wavelet with optimized parameters.The wavelet coefficients containing impact component are selected using kurtosis as an index.Finally,the impact signal are reconstructed using selected coefficients by inverse wavelet transform,so that the noise suppression and weak signal detection and extraction can be realized.Results of simulation experiment show that this method can effectively suppress noise and still outperforms other methods at low SNR.Aiming at the problem of separation of multi-source mixed fault signals,a source separation method based on Independent Component Analysis(ICA)is proposed.Firstly,the principles of Blind Source Separation(BSS)and independent component analysis are elaborated.Then,the feasibility of applying independent component analysis to signal separation of mechanical fault sources is discussed.FastICA algorithm based on the symmetry orthogonalization based on negative entropy for estimating multiple independent components is used to separate the vibration signals collected by multi-acceleration sensors.Resonance demodulation is then performed on the obtained several groups of independent components.At last,independent components containing fault features are confirmed by spectrum analysis,and the impact signals with specific fault features are separated.By analyzing and processing the simulation signal and bearing pitting fault signal,it is verified that the proposed method can separate the fault signal from the mixed signal of multi-sensor.Aiming at the problem of locating the fault source of weak impact signal,a method of determining the starting point of impact signal based on Hilbert-Huang Transform(HHT)and Akaike Information Criteria(AIC)is proposed.The signal is decomposed into several modes by Empirical Mode Decomposition(EMD).Then the Hilbert transform is performed to obtain the instantaneous frequency of each mode and the minimum of AIC equation is searched,and the corresponding sampling point is the starting point of the impact signal.Simulation results show that the proposed method can accurately determin the starting point of the impact signal,and still maintain good accuracy and stability as the background noise increases.Aiming at the feature extraction of periodic impact signal of rotating machinery,a method of extracting the frequency feature of impact signal based on mathematical morphology filtering is proposed.Two parameters,namely the length of structural element and the scale of morphological filter are two-dimensionally optimized by taking the corresponding amplitude of fault characteristic frequency after morphological filtering as the index,and the optimal morphological filter is obtained.Through the processing and analysis of the simulated impact signal and the real bearing pitting fault signal,it is verified that this method can extract the frequency characteristics of the periodic impact signal to the maximum extent under strong background noise.