Research on ultrasonic non-destructive evaluation algorithm for ultimate tensile strength of stainless-steel welded pipe welds based on width learning

Abstract

ABSTRACT The regression prediction between ultrasonic signal and ultimate tensile strength of stainless-steel pipe weld is studied in this paper. First, ultrasonic signals were collected from the weld of AISI304 austenitic stainless steel welded pipe sample, and its ultimate tensile strength was tested. The data sets of ultrasonic signal and ultimate tensile strength of weld were constructed. The random forest algorithm is used to extract the characteristic value set of the ultrasonic signal set of the welding seam and reduce the dimension of the characteristic value set. Then, the set of eigenvalues after dimensionality reduction is input into the Bayesian BLS network for training. The results of the training were compared with those of other regression models. The results show that the model established by Bayes BLS algorithm is feasible and reliable to predict the ultimate tensile strength of stainless-steel pipe weld. The results also show that the model established by the Bayesian BLS algorithm has high accuracy and stability in predicting the ultimate tensile strength of stainless-steel pipe weld. This method can be used to predict the ultimate tensile strength of welds in different types of materials.