Local type fault diagnosis method of rotating machine based on sparse decomposition optimization algorithm

Abstract

The invention discloses a local type fault diagnosis method of a rotating machine based on a sparse decomposition optimization algorithm, which comprises the following steps: S1, acquiring a rotatingmachine vibration acceleration response signal containing fault characteristic information; S2, intercepting a signal with an appropriate length, performing noise reduction pretreatment by applying ahigh-pass filter and singular value decomposition, and improving the signal-to-noise ratio; S3, setting an atom optimization criterion, and establishing a time-frequency domain correlation coefficientconstraint function of atoms and an actually-measured signal; S4, updating parameters of an initial random dictionary by using a particle swarm hybrid gradient descent algorithm to obtain an optimized impact response dictionary; S5, solving a sparse coefficient by using a segmented Lagrange contraction algorithm, reconstructing the fault characteristic signal according to the coefficient and thedictionary; S6, analyzing the impact response time interval of the reconstructed signal and demodulation of the characteristic frequency of the amplitude spectrum, identifying the position of the fault to finish the fault diagnosis. The dictionary has the advantages of high precision, high speed, stronger noise resistance and more accurate reconstruction characteristic signal.

Description

technical field [0001] The invention belongs to the field of mechanical fault diagnosis, and specifically relates to a partial fault diagnosis method for rotating machinery based on sparse decomposition and using a particle swarm hybrid gradient descent optimization algorithm, which can be used to extract fault characteristic signals of rotating machinery such as bearings and gearboxes, and perform Troubleshooting. Background technique [0002] Rotating parts such as bearings and gears are an essential general-purpose component in mechanical equipment and one of the most easily damaged elements. Fault diagnosis for them is of great significance for prolonging the service life of equipment and reducing maintenance costs. How to extract the characteristic signal containing fault information from the noise signal is the key to diagnosis. [0003] Commonly used fault feature signal extraction methods mainly include amplitude demodulation based on Fourier transform, wavelet tra...

AI Technical Summary

Problems solved by technology

The traditional learning dictionary is obtained by machine learning, which has a high degree of matching with the measured signal, but the calculation is heavy, and the dictionary lacks clear physical meaning; while the analytical dictionary has the advantages of mature theory, but it ignores the differences between different measured signals. difference, matching adaptability is not good enough

Images