Defect Feature Extraction And Intelligent Recognition Of Ultrasonic Detection Signals In Burrs Of Welded

The pipelines need strong anti-collapse ability and low cost in the process of oil and gas transportation.In oil,natural gas and other industries,the ERW is used mostly widely in this range.In order to prevent the potential hazards better and extend the life of welded pipe,we must be more strict in the quality in industrial production.With the rapid development of technology,ultrasonic testing technology is more and more widely used in the field of nondestructive testing.The fully using of the signal information is difficult due to the lower time-frequency resolution,which leads to a poor detection accuracy and reliability.It can be seen that the local features of time-frequency analysis results are more effective in describing the signal characteristics,and it is more effective for the analysis,recognition of ultrasonic signals,the improvement of detection accuracy and reliability.In the practical ultrasonic testing of the inner burr,it can not be 100% to realize the qualitative classification of a certain defect,therefor it is necessary to continuously explore and research in order to achieve the feature extraction of the defect.Based on the point,this paper uses MATLAB software to analyze the time-frequency localization of the ultrasonic signal of the burr,and extracts the features of the defect signal for laying a solid foundation for the detection of the burr in the future.In order to overcome the problem of modal aliasing in the traditional EMD method,it is necessary to add a large number of Gaussian white noise before processing the signal that named as the EEMD method,which greatly reduces the speed of the EEMD.Orthogonal wavelet packet as a pre-filter unit of EEMD method is effective to improve the timeliness.During the actual process of removing the inner burr,there are various types of burrs in the weld due to the different position and the usage time of scraper,and the ultrasonic test results of these burrs are also quite different.According to amplitude characteristics and thickness characteristics can determine whether the burrs and burrs types.As the beam directivity of ultrasonic probe in the liquid is poor,and there are many effects of interfering waves.Using the center frequency of 2MHz water immersion line focusing probe to collect sample data of the ultrasonic signal was sampled from ERW welded pipe with inner burr,which provide real and reliable data.And it is applied to the feature extraction and intelligent recognition of defects.By observing the wave shape of the ultrasonic echo signal of the burr in the pipe,and known the number of sampling points of defect in the waveform,the time of the single sampling point,the propagation speed of the ultrasonic wave in the medium and the incidence angle of the ultrasonic probe,thenthe location and depth of the burr would be accurately obtained.Because some effective information of ultrasonic echo signals are submerged in a large amount of noise,EEMD method with high time-frequency resolution is effective to decompose the multi-scale signals,and the obtained results can fully reflect the information characteristics of the original signal.And the system parameters of the multi-scale IMF signals are predicted by the Volterra series prediction model based on Lorenz chaotic system.The least squares solution of the system is obtained by calculating the matrix singular value,and the prediction precision is improved.According to the size of the obtained singular value can effectively determine whether the existence of the inner burr,which verify the correctness and effectiveness of EEMD-Volterra method on the inner burr detection in this paper.