An Improved Resonance-based Sparse Decomposition Algorithm And Its Application On The Fault Diagnosis Of Mechanical

To date,the development of industry tends to be more automated,large-scaled and intelligent.Therefore,the mechanical driving mechanism has been widely used in large mechanical equipment as a part of the transmission.Rolling bearing plays a key importance in the large-scale machinery equipment.When rolling bearing fails,it may cause huge economic losses and out of normal conditions about the mechanical equipment.Thus,it is necessary to pay attention on the running state.As the feature in the early fault is too weak to hardly been recognized during the diagnose,an improved method of resonance sparsity decomposition is introduced in the fault diagnosis of rolling bearing.The effectiveness of the method is verified by applying the proposed method to the simulation signal and measured vibratio n signal in mechanical equipment.1.The principle theory of resonance sparsity decomposition is represented in the chapter.As the method regards the vibration signal as a sum of periodic harmonic signal and transient pulse signal,the signal will be deco mposed as high resonance component and low resonance component,according to the characteristic of this theory,the resonance sparsity decomposition is applied to extract the pulse signal,simulation signal and measured signal are used to validate the effectiveness of it.2.Aiming at the effectiveness of decomposition is effected by the parameters selected freedom,an improved resonance sparsity decomposition based on Particles swarm optimization is proposed here with the characteristic of global optimizat ion,correlated kurtosis is also applied here to evaluate the amount of pulse in the low resonance component.Finally,the effectiveness of the proposed method is validated by the decomposition of Simulation signal and measured signal,moreover,variationa l mode decomposition is here to compare with improved resonance sparsity decomposition.3.Resonance sparsity decomposition is used to perform the classification for rolling bearing signal and gear signal.In order to have a good classification of mechanical fault,the coefficients' Permutation entrop y are utilized as feature vector to conduct the classification with KNN classifier.The effectiveness is validated by comparing with wavelet.