A Study Of Laser Ultrasonic Damage Signal Processing Methods Based On DEMD

With the development of the science technology, laser ultrasonic gradually becomes themain test method to detect the surfaces of complex equipments or micro damages of thoseinternal parts, whose features are non-contact, small light spot, high resolution and mobilelight, etc. In this paper, a series of studies are conducted to do some quantitative analysisand feature extractions of micro injuries of aluminum surfaces.Firstly, the theories of laser ultrasonic tests of surface damages are discussed and themethods and the specific structures of the experimental system are described by analyzing thecharacteristics of the damage signals. Regarding the experimental echo signals asnon-stationary and nonlinear stochastic signals, the methods based on EMD (Empirical ModeDecomposition) are used to analyze and process the damage signals. As the traditional EMDhas an end effect, DEMD (Derivative-optimized empirical mode decomposition) is adopted.DEMD algorithm improves the traditional cubic spline interpolation fitting for the upper andlower envelope, which is better than the original one, for its IMFs (Intrinsic Mode Function)have higher correlation, lower mean square error and better suppression of the end effects.Secondly, laser line source heat-structure simulation models are established based onthe mechanism of excitation source for thermoelastic excitation. DEMD is adopted todecompose the signals of the numerical simulation and analyze the time-frequency domainand instantaneous frequency of the IMFs. The results show that there are obvious variationrules between the time-domain waveform and the instantaneous frequency of the IMFs andthe crack depths.Finally, damage echo signals which are obtained through the laser ultrasonic surfacedamage detection experimental system are pretreated by DEMD based on the formula ofsignal-noise modal sorting. In accordance with Hilbert spectrum and marginal spectrumtheory, the experimental signals are DEMD decomposed; and the instantaneous frequencyanalysis, Hilbert marginal spectrum analysis and spectral analysis are carried out to detect thescreened IMFS separately. It turns out that the results obtained have higher recognition rateand more accurate eigenvalues, which can provide quantitative reference standards to judge different crack depths and some reference to detect the surface damage when using laserultrasonic.