Comparative Research On Adaptive Signal Decomposition Algorithm And Its Application In Hydraulic Pump Fault Diagnosis

Adaptive signal decomposition algorithm is a powerful tool for signal analysis,because it can decompose a signal into multiple narrowband components,which is beneficial to quantitatively evaluate signal characteristics and has an important impact on the reliability and accuracy of fault diagnosis.The hydraulic pump is the core component of the entire hydraulic system,and the failure detection of the hydraulic pump is mostly realized through the analysis of vibration signals,pressure signals and other data.Signal processing is an important part of the fault diagnosis process,and the quality of its processing results will directly affect the correct rate of the subsequent pattern recognition process,so it is of great significance to apply the adaptive signal decomposition algorithm to the fault diagnosis of hydraulic pumps.In this article,five adaptive signal decomposition algorithms are divided into two types based on time domain and frequency domain.Among them,five adaptive signal decomposition algorithms include Empirical Mode Decomposition(EMD),Ensemble Empirical Mode Decomposition(EEMD),Adaptive Local Iterative Filtering(ALIF),Empirical Wavelet Transform(EWT),Variational Mode Decomposition(VMD).In the second and third chapters,the differences in their decomposition principles,the advantages and disadvantages of the algorithm implementation,and some simulation signals and actual bearing signals of Western Reserve University are provided to illustrate the decomposition effect of the algorithm and the main factors affecting the effect of the algorithm..In Chapter4,for the hydraulic pump loose shoe failure,sliding shoe failure,center spring failure,swash plate wear and normal state signals collected by the experimental platform,first use EEMD and EWT algorithms to denoise the vibration signal,and then reconstruct The signal undergoes envelope spectrum analysis.Since EWT has the problem of incomplete reconstruction due to frequency band division,the information contained in the envelope spectrum of the final EEMD reconstructed signal is better than EWT.In the fifth chapter,three methods of EEMD,ALIF and VMD are used to denoise the five state signals,and then the skewness,kurtosis,permutation entropy,center of gravity frequency and wavelet packet are extracted in the time domain,frequency domain and time-frequency domain.The normalized energy is used as the feature to construct the feature vector.Finally,through the results of Support Vector Machines(SVM)fault classification,it is concluded that VMD is the most suitable for the denoising of hydraulic pumps among the three methods.