| Acoustic emission(AE)is an important approach for no-destructive measurement and evaluation,which is widely used in a number of fields,such as petrochemical industry,aerospace industry,material test,and transportation.The operation safety and reliability of large industrial infrastructure are hot topics in the current research field.If the performance and useful lifespan of the industrial infrastructure were not effectively detected and evaluated,catastrophic accidents would be hard to avoid.The key technology of AE detection is researched in this work,the contents of the work are listed below:(1)Based on the commonness of AE signals,the propagation,attenuation and scattering models of AE signals,and the influence of noise on fractal dimension of AE signal are analyzed.By adopting the pattern of Fast ICA noise reduction,the effects of four different criterions of FastICA model on noise reduction are analyzed and compared.The results show that the FastICA noise reduction effect is better than other,and the AE characteristic parameters of the signal processed by FastICA noise reduction are compared with the original one,and the feasibility and validity of the FastICA algorithm in noise reduction of AE signal are verified.(2)Aiming at the problem of uncertainty in the detection of AE event,this paper adopts the method of AE signal processing based on clustering analysis,which improves the accuracy of AE event extraction.According to the requirement of AE localization,the triangle positioning algorithm and the ultra-fixed positioning algorithm are researched,and due to the difference of the two algorithms,the Newton iterative method and trust region method are used respectively to solve the positioning equation.The computational performance and positioning performance of two algorithms in ideal and non-ideal conditions are studied by Monte Carlo simulation experiment,and the effects of velocity error and time error on two algorithms are analyzed.The results of AEWin,triangle positioning and ultra-positioning are compared through the lead-breaking experiment,and the results show that the algorithm based on hyper-positioning presents higher positioning accuracy and better stability.(3)The bending experiment of two kinds of Q235B plate specimens was designed.The whole time frequency process and correlation analysis were used to analyze the characteristic parameters of AE signal during the damage process.Through the mechanism analysis can be obtained,when the material has defect,AE signal activity will increase more rapidly,more easily to reach the yield point,and with the occurrence of new characteristic frequency.the correlation between AE characteristic parameters is only sensitive to AE source,which is irrelevant to the structural damage form that generates the AE signal.Based on the AR model,the AE signal is extracted from the Q235 stretching process,and the model parameters of AR modeling are used as the input of BP neural network to realize the identification of the specimen at different damage stages.Compared other algorithms,the result shows that pattern recognition method based on AR and BPNN has higher recognition rate.(4)A set of "AE detection data analysis system" software was independently developed.Its main functions include:AE signal acquisition module,signal Processing module,display module and document preservation module.This system can realize the waveform analysis of AE signal,AE source localization and pattern recognition. |
PDF Full Text Request