Research On Nondestructive Testing Of Electromagnetically Induced Acoustic Emission Based On Wavelet Morphology And Neural Networks

Electromagnetic induced acoustic emission technique is a novel non-destructive testingwith electromagnetic loading acoustic emission signal, which gathers the advantages ofelectromagnetic and acoustic emission technologies. According to the needs under differentsituations, any area of EUT can be detected by this method at any time. Not only does notcause unnecessary damage for the component integral, but also shortens the load time ofcommon acoustic emission testing. The main research work is as follows:This paper studies the mechanism of electromagnetic loading and stimulating acousticemission signal. By using ANSYS finite element analysis software, the defect and zero defectmodels has been established are loaded the same current value. Meanwhile, with changing theinput current value, the current distribution, magnetic flux density distribution, electromagnetic forceand crack deformation of the two models are analyzed comparatively. Hence, the loadingmechanism is proved by the analysis result of metal plate stimulating acoustic emissionsignal under the heavy current.The appropriate wavelet basis function of signal is selected when analyzing theelectromagnetic acoustic emission signal. The signal are decomposed by using wavelet packet,and which of relevant characteristics are extracted to ensure the energy characteristicscriterion of the signal. In addition, during the experimental phase, the noise interference ofelectromagnetic acoustic emission signals should be considered fully. The de-noisingprogram of integrating wavelet decomposition and morphology is designed based on the waveformcharacteristics.The wavelet theory and neural network are combined to apply to electromagneticacoustic emission signals recognition, which can identify the metal plates whether havecracks or not. The wavelet neural network is improved to shorten the training time and raiserecognition accuracy. Furthermore, by analyzing the signal energy feature criterion, thenetwork characteristic is enhanced with optimizing the signal characteristic parameters.