Design Of Adaptive Sparse Dictionary And Its Application On Hybrid Fault Diagnosis Of Gear System

Gears and bearings are both important components in modern industrial machinery.The long time service of gears and bearings under complex and time-varying,harsh environment leads to various types of faults,even potential safety hazards.How to quickly identify faulty components from signals containing a large amount of noise interferences and effectively conduct fault diagnosis and equipment monitoring are of great significance for ensuring equipment and personal safety.Based on the sparse representation,the research focuses on the extraction of fault characteristic signals of bearings and gears.A new method for extracting the impact fault signal of rolling bearings is proposed.Considering the characteristics of impulse response in the time-frequency domain,the analytical dictionary is linked with the measured signal to build a new impulse response atomic constraint function.The obtained atoms equips with clear physical meaning and high matching degree with the actual signal,which could better represent faulty characteristics.The proposed method combines the advantages of Particle Swarm Optimization and Gradient Descent algorithm to accelerate the solution to the constraint functions.Different update strategies are adopted for particles with different fitness intervals in the fusion algorithm to balance the convergence speed and accuracy of the solution.There is a greater speed advantage for signals at high sampling frequency.At the same time,singular value decomposition is introduced to preprocess the signal,which reduces noise and highlights the characteristic signals,as well as improves the precision of atoms and coefficients.The simulation results coincide with the experimental results,indicating that the proposed method is not only faster than the contrast method but also with higher precision,stronger anti-noise and better adaptability.In short,it can better extract fault characteristic signals of the inner and outer rings of the rolling bearing,and effectively defect fault.Regarding to the gearbox system with more complex signal components than bearing systems,an extraction method for coupled modulated vibration signals of hybrid faults is proposed.The interpolation correction method is used to correct the atomic parameters in the steady modulation dictionary to get higher precision.The separation of the coupled faulty signals is completed by two steps of sparse decompositions.The effect of three important parameters of SALSA on separation results are researched separately.Besides,the clear selection criterions of atomic length,regular term and penalty parameter are adaptively selected according to the energy ratio operator,which could reduce the mistakes and errors of manual selection.The simulation verifies the effective of extraction under different SNRs.The results show that even though the accuracy of the atomic parameters is dropped a little when the SNR of hybrid faulty signal is-3dB,-6dB,-9dB,all the impact signals are still reconstructed well.Both the steady modulation signal and the impact response in the experimental signals of the fixed shaft gearbox and the planetary gearbox are successfully separated.The relative error of the impact faulty period is less than 1%,which further proves the effectiveness of the proposed method.