Identification Of Crack Propagation Mode Of Fatigue Specimen Based On Acoustic Emission Signal

The study of fatigue crack propagation state is the basis of research on fatigue crack characteristics of materials.The acoustic emission technology realizes the recognition of the fatigue extended state by continuously monitoring and analyzing the acoustic emission signal of the received fatigue test piece during the crack propagation process.However,due to the large amount of noise in the acoustic emission signal,the characteristics of the acoustic emission source cannot be directly obtained,resulting in low accuracy of crack identification.At the same time,how to select the best feature parameters and select the optimal reasoning method in the fusion information directly affects the accuracy of the identification.Therefore,in order to more accurately determine the crack propagation process of the fatigue specimen under laboratory conditions,It is necessary to identify the real-time extension state of fatigue specimens and to study the processing technology based on acoustic emission signals and pattern recognition technology.Based on the acoustic emission signal,this paper studies the signal denoising,feature extraction,information fusion,feature screening and pattern recognition methods in the crack pattern analysis of fatigue specimens,and combines simulation and experiment methods.The comparative study determines the specific methods of each processing link of acoustic emission signals.The validity and practicability of the pattern recognition method is verified by the pattern recognition of the crack propagation state of Q235 central crack tensile specimens.The specific content of the topic is as follows:(1)Through the research on the denoising technology of acoustic emission signals,comparative analysis of wavelet denoising,wavelet packet transform denoising,empirical mode decomposition denoising and other denoising methods,an acoustic emission signal is proposed.The effectiveness of the proposed method is verified by simulation signal and experimental signal analysis;(2)In the feature extraction,the principle and process of wavelet packet energy spectrum and singular value decomposition result are analyzed as feature parameters,and the effectiveness of the algorithm is verified by extracting the features of simulated acoustic emission signal and broken lead acoustic emission;(3)In the information fusion and parameter screening,summarize the basic theory of information fusion and determine the final information fusion method.For the obtained fusion feature set,use Principal Component Analysis and Kernel Principal Component Analysis to filter the parameters.Analyze the screening effects of the two methods and their applicability to the subject;(4)Based on the research on the commonly used pattern recognition methods,the principles and implementation processes of BP neural network,RBF neural network,common multi-classification support vector machine and multi-classification support vector machine based on least squares are expounded and determine the final pattern recognition scheme of the system;(5)Through the pattern recognition of the crack propagation state of the Q235 center crack tensile specimen,it is verified whether the acoustic emission signal denoising method proposed in this paper is effective and reliable,and whether the information fusion method can complete the information fusion of the two-channel acoustic emission signal,whether the characteristic parameters after dimensionality reduction have good characterization ability for the crack propagation state of fatigue specimens,and the recognition effects of various pattern recognition methods determined by comparative analysis under the conditions of this subject;(6)Based on the Matlab software platform,the function of the acoustic emission signal pattern recognition system for fatigue test pieces is completed.