The Ultrasonic Weld Defects Detection And Recognition Based On Sequence Alignment Algorithm

With the continuous development of modern industry,welded components has been widely used in areas such as aerospace,railway and energy,however,the weld detection is necessary to ensure the safety.Currently,the conventional weld defect detection mainly uses wavelet packet to decompose or analyze the energy change of ultrasonic echo signal to extract defect characteristics and uses support vector machine(SVM)or the image segmentation method to build the defect recognition classifier,certain results has been achieved in theory and application.These methods generally fixed constructed defect classification model and on the finite quantity training data set,but the ultrasonic weld defect data with the characteristics of large-scale,extensible,and explored is increasing as the production test proceeding,the conventional defects recognition method is difficult to reflect the polymorphism of similar defects,the fixed defect classification model is unfavorable to fill the new data.To solve the problems of the conventional ultrasonic weld defect recognition method is difficult to reflect the differences of different defects and the polymorphism of similar defects with fixed classification model and the limited size of the training set,combining with the data analysis strategy and genetic defect recognition matching mind to identify the defect.But the defect feature sequence and the genetic sequences are very different,so convert the defect echo signal,analysis a large number of historical testing characteristic values with the principal component analysis and cluster analysis,extract the main characteristics of the defects and got the weight of each feature and its characteristics of values,convert the defect detection data to comparable object,then the current detection of defect features could match with the historical testing data and identify the porosity,slag inclusion,cracks,incomplete penetration and incomplete fusion five kinds of common defects.At the same time,build the perfect model of the history testing data set on the basis of the nearest neighbor method,add the latest testing data with difference choice,expanded the tolerance range of the defect features in the history testing data set,ensure the alignment accuracy.Based on the sequence alignment algorithm,a new method of defect recognition in ultrasonic detection is studied.This method gets the results of the porosity,slag inclusion,cracks,incomplete penetration and incomplete fusion five kinds of history defect feature by the matching between the characteristic values extracted from the current inspection data and the historical data sets,and identify the type of the current detecting defect.Through the experiment verify the method is effective with the high recognition accuracy,improves the utilization rate of the defect information and avoid the problem of difficult to reflect the polymorphisms of the similar defects in the conventional defect detection method,improves the reliability of the defect recognition further.