A De-noising Method Based On Common Components Reconstruction Of Rotating Machinery Homologous Responses

Condition monitoring and fault diagnosis are of great significance to ensure the safe and reliable operation of key equipment.The original vibration data often contains the bad data caused by the sensor and is affected by the uncertainty of the working condition,which is the main factor of false alarm in the state monitoring system.Screening the original vibration data can eliminate the influence of inferior data and reduce the probability of false alarm.This paper proposes a denoising method which extract and reconstruct common components to obtain real response signals.Analyze the performance of different decomposition methods to different types of noise.The analysis of signal decomposition performance is the basis of common components extraction de-noising and the guarantee of finding the common component from homologous fault multiple response.Signal decomposition methods based on different decomposition principles(wavelet transform,EEMD decomposition,SSA decomposition and resonance-based sparse decomposition)are applied to decompose the impulse responses interfered by local noise and white Gaussian noise respectively,as well as multi-frequency component harmonic signal.Analyze the performance of different decomposition methods in terms of the energy distribution,SNR and other angles of decomposed subcomponents,fromVerify the feasibility of extracting common components reconstruction with wavelet decomposition.The multiresolution property of wavelet decomposes the multiple responses of homologous fault.Decompose the multiple responses of homologous fault by multi-resolution wavelet transform,measure the similarity of decomposed subsignals,determine the threshold,extract and reconstruct the common components to complete de-noising.In the identification of common components,the Dynamic Time Warping algorithm is selected as the time series similarity measurement method,and K-means clustering is selected as the automatic threshold determination method.The extracted common components de-noising based on wavelet decomposes the simulation impulse bearing signal,the measured bearing signal with gear meshing frequency interference and variable speed,and the feasibility of proposed de-noising method is proved by experimental results.The denoising method of wavelet-SSA double layer common components reconstruction is proposed.The simulation signals verify that SSA decomposition can make up for the decomposition insufficient of wavelet transform dealing complex frequency component signals.Due to the limitation of wavelet filter principle,it is possible for wavelet decomposition extract common components mix interference noise in common component signal or omit common components in residual noise.Using double-layer de-noising can find out more real common components to make up for this deficiency and improve the de-noising effect.The denoising of simulated signal and measured signal indicates the feasibility of the wavelet-SSA double-layer extraction common components reconstruction denoising.Aiming at the separation problem of mixed signals with overlapping frequency bands,a de-noising method for common components reconstruction by resonance-based sparse decomposition is proposed.The frequency band aliasing bearing signal with inner race fault and outer race fault are sliced according to the characteristic frequency of inner race fault.resonance-based sparse decomposition is used to decompose the multiple mixed responses into high resonant components and low resonant components.Through the similarity measurement of the decomposed sub-components,find the common components representing the inner race fault,and separate the inner race fault response under the interference of the outer race fault.The signal reconstructed by common components highlights the characteristics of the bearing inner race fault,and the envelope spectrum can remove the interference of the outer race fault characteristic frequency,showing the clear inner race fault characteristic frequency,and proving the feasibility of the de-noising method.Based on the above research,this paper summarizes rotating machinery de-noising method for the extraction common components from the homologous multiple responses.Wavelet transform,SSA and resonance-based sparse decomposition method are used as the technical means to complete the denoising tasks for different types of noise-containing signals of rotating machinery.