Research On Recognition Of Defects Of Welding Lines In The Ultrasonic Phased Array Testing

With the development of computer artificial intelligence technology and electronic technology,field of ultrasonic testing has reached the new era.As a new technology in the field of ultrasonic testing,compared with traditional ultrasonic technology,ultrasonic phased array not only has the advantage of high sensitivity,strong penetrability,operations security,a wide range of application,but it also has its own the advantage of operating conveniently and creating high-resolution imaging.As a result of the shot time,the application of the ultrasonic phased array to NDT(Non destructive testing),it lacks effective recognition method of the test for the butt weldin g defect type.In order to researched the influence of different the number of probe array element and design a high recognition rate of defect recognition of program,a method based on GGCM(or GLCM)and grey correlation degree is presented.In order to control the number of using probe array element,a procedure based on Visual Studio 9.0 and MATLAB 10.0 is presented.In experiment,different thickness of weld defects are tested by ultrasonic phased array in two conditions,that is,the number of probe array element being 15 and 32.Then the experiment has gain 37 groups of defect images and can create a database.experiment makes a preliminary analysis of the influence in two conditions by image matching testing.In order to analyze 5 types of weld defects(crack,slag,porosity,incomplete penetration,incomplete fusion),a texture feature extraction method based on GLCM can be used.In order to identify the different types of ultrasonic phased array inspection imaging of weld defects,four directions(0°?450°?90°?135°)of image characteristics(ASM,CON,H,COR)based on GLCM and 15 characteristic parameters based on GGCM are separately combine with gray relative analysis method.According to recognition results,This paper presents an overview of GGCM and GLCM of defect recognition methods,as well as the methods strong points and short comings.Therefore,it also show the influence of different the number of probe array element.Finally,experimental result showed that the quality of image in 32 probe array element is beter than its in15 probe array element and the result of image matching testing is correct.Besides,the results of this two kind of defect recognition methods are good,both results higher than 80%.However,But due to the huge workload and the complexity of calculation and lower rate of between incomplete penetration and incomplete fusion,defect recognition method based on GLCM is worse than that based on GGCM.