A Study On Ultrasonic TOFD-D Scanning Imaging Technology

The welding structure as an important metal structure had been widely used in modern industry.However,it has still been limited by the disadvantages of welding technology.There are a variety of defects which arise in the weld joint.It is necessary to do nondestructive examination for the weld joint to guarantee the safety of the welding structure.Time of Flight Diffraction(TOFD)is a conventional method for detecting weld defects.The method has the advantages of high detection speed,low cost and harmless to the human body,which can be used to evaluate defects quantitatively.Ultrasonic TOFD-D scanning characterization and automatic recognition of typical defects of butt welds are studied in this thesis.Firstly,ultrasonic TOFD detection process was simulated through finite element method(FEM).Depending on the FEM results,ultrasonic propagation mechanism and the influences on testing signal were observed.A large number of samples of butt weld are prepared and are scanned by ultrasonic TOFD-D.Different features of every kind of typical weld defects is analyzed.Finally,according to the characteristics of the typical defect of TOFD-D scanning image of the weld,the design of defect automatic recognition classification algorithm is provided and lots of training and identification of sample is obtained by changing the testing parameters to train the classification network.The specific research contents are as follows:Firstly,finite element method is used to simulate the ultrasonic TOFD detection process of butt weld in this research.Different forms of circular holes,slot,deep groove defects in weld is simulated and the influence of the defect type,depth,size and surface defect orientation on the detection signal amplitude is analyzed.The simulation results show that the finite element simulation can well simulate the formation mechanism of the diffraction wave of the volume and the area type defect.It is found that the distribution characteristics of the defect diffraction wave indicate a certain correspondence with the shape of the defect.Secondly,the ultrasonic TOFD-D scanning technique is used to image steel welds with typical defects,and the imaging features of various defects are analyzed.By comparing the scanning images,it is found that the feature of the scanning image are parabolic,linear and so on,which can be used to distinguish the defect’s type are mainly the opening direction of the parabola,whether the shape is continuous or not,and whether the end point is smooth.Finally,an automatic recognition and classification algorithm of defects in ultrasonic TOFD-D scanning image which are based on Faster R-CNN algorithm are designed in the thesis and a high rate of automatic recognition is obtained after that.The Faster RCNN algorithm is optimized for the characteristics of ultrasonic TOFD-D scanning imaging.In the training phase,the expansion of data is used to avoid under-fitting,and enhance the recognition system of anti-noise ability and robustness.The feasibility,accuracy and efficiency of the method which is applied to ultrasonic TOFD-D imaging were analyzed through experiments.It can be conclude that the detection signal and scanning image of ultrasonic TOFD-D are closely related to the formation of defects.The ultrasonic TOFD-D scanning image can be used to distinguish the defect’s type.The automatic recognition algorithm of defects’ type can be applied to ultrasonic TOFD detection,which can realize the rapid and accurate detection of defects.The study of automatic identification method of typical defects in weld has great significance for the detection automation.