Font Size: a A A

Finite Element Analysis And Signal Processing Of Electromagnetically Excited Acoustic Emissio

Posted on:2024-07-22Degree:MasterType:Thesis
Country:ChinaCandidate:C L YangFull Text:PDF
GTID:2531307052465674Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
Electromagnetic acoustic emission technology(EMAE)is a kind of non-destructive testing technology.This technology can be used to detect crack defects of metallic materials by loading them with excitation to excite acoustic emission signal.It is mainly used for on-line monitoring of metal structures such as pressure vessels,impellers and aircraft skins.Based on this,magnesium alloy AZ31 B was used as a research object in this paper,and round holes and cracks were used to simulate the crack defects around rivet holes or bolt holes which can be seen frequently in real life,to study the magnesium alloy specimens with crack defects.Simulation and experimental research were conducted on the key issues such as the mechanism of electromagnetic acoustic emission phenomenon,feature extraction and identification of electromagnetic acoustic emission signal.Specific work includes:(1)The electromagnetic acoustic emission finite element models based on non-contact loading of eddy current coil and high current contact loading were established for round hole and cracked magnesium alloy specimens respectively by using COMSOL software to verify the feasibility of electromagnetic acoustic emission technology in the detection of crack defects in magnesium alloy specimens.By comparing and analyzing the two loading methods,it is concluded that the eddy current coil excitation method is more feasible in practice.Therefore,a receiving model of the EMAE simulated signal was established,and non-stationary,non-linear burst type EMAE signal was acquired,providing theoretical evidence for the detection mechanism of electromagnetic acoustic emission.Based on the finite element analysis,an experimental platform for eddy current loaded EMAE was built.The experimental signals of three kinds of magnesium alloy metal specimens,including round hole and crack,round hole,and intact,were collected,and the non-stationary and non-linear burst acoustic emission signals were obtained,which is consistent with the acoustic emission signal collected by the finite element analysis.The validity and correctness of the finite element analysis is further demonstrated experimentally.(2)To address the shortcomings of conventional time-frequency analysis methods in the processing of electromagnetic acoustic emission signals,such as inflexible wavelet basis functions or cross term interference,a method of Feature extraction and signal recognition of EMAE signals based on the combination of generalized S-transform and Extreme Learning Machine(ELM)is proposed.Firstly,the time-frequency analysis of the EMAE experimental signal is performed by using the generalized S transformation.It is found that the energy of the time-frequency diagram of round hole and crack magnesium alloy specimens is distributed in low frequency and high frequency parts.However,the energy of the time-frequency diagram of the round hole magnesium alloy specimen and the intact magnesium alloy specimen is mainly distributed in the low frequency part.Secondly,to further extract the features that can characterize the EMAE signal,the mode-time frequency matrix obtained after generalized S-transform is partitioned by frequency into high and low frequency bands,and Tsallis singular entropy values are calculated to construct a two-dimensional feature vector.Finally,The feature vectors are trained and tested using ELM.The results showed that the recognition accuracy of this method can reach 92.5%.This method can effectively identify the electromagnetic acoustic emission signal.(3)In order to further improve the recognition accuracy of EMAE signal,and to address the shortcomings that it is difficult to accurately extract the feature information of non-stationary non-linear signals,a method combining local mean decomposition(LMD)with least squares support vector machine(LSSVM)is proposed for the identification of EMAE signals.Firstly,non-stationary acoustic emission signals are decomposed into several stationary product function by local mean decomposition.Secondly,in order to reduce training time and improve recognition efficiency,the best PF component signal is selected by using the energy proportion method,and AR model is established.The AIC criterion is used to determine that the best order of AR model is 5th order.The parameters and mean square deviation of the model are extracted as signal characteristics,and a six-dimensional eigenvector is obtained.Finally,the LSSVM classifier optimized by cross-validation is trained and tested to identify the EMAE signal.The results show that the recognition accuracy of this method can reach 97.5%,which is higher than that of SVM and BP network.
Keywords/Search Tags:Finite Element Analysis, Generalized S Transform, Extreme Learning Machine, Local Mean Decomposition, Least Squares Support Vector Machine
PDF Full Text Request
Related items