Application Research for Ultrasonic Flaw Identification Based on Support Vector Machine

Abstract

Listen

Translate article

Automatic identification of flaws is very important for ultrasonic nondestructive testing and evaluation of pipelines. A novel automatic identification approach of flaws using support vector machine (SVM) is presented. Wavelet transform is applied to feature extraction of ultrasonic echo signals, and SVM is to perform the identification task. To validate this approach, some experiments are performed. The results show that unlike conventional and artificial neural networks (ANN) identification methods the new technique performs better than conventional evaluation ones with advantages of high identification performance for pipeline flaws, lower cost, excellent generalization. Introduction Pipelines prove to be the safest and the most economical means transporting large quantities of oil and gas resources. Maintenance of oil and gas pipelines is an issue of great concern for all oil and gas companies. But various flaws will inevitably occur and grow during the operations, such as crack, leakage, corrosion of the pipe wall, etc. Flaw detection and identification is a very important step to ensure pipelines safe operation. The traditional flaw detection suffers from complicated process, low accurate rate and off-line implement. The improved methods of flaw identification by artificial neural networks (ANN) can lead to the problems of overfit and bad generalization because of finite samples[1]. In this paper after de-noising the ultrasonic echo signals using wavelet transform and with a view of data mining, a novel approach using SVM classification is discussed to identify the flaws. The experiment results show that unlike conventional and ANN identification methods the new technique performs better than conventional evaluation ones with advantages of high efficiency, lower cost, easy implement on-line, excellent generalization. The approach provides a novel technique means for nondestructive flaw identification of various flaws. Pipeline flaw identification based on SVM A. Theory and principle of SVM SVM initially came into prominence in the area of hand-written character recognition and is now being rapidly applied to many other fields, such as text categorization, computer vision, speech recognition and gene classification, etc.[2-3]. SVM is the approximate realization of the structural risk minimization method, analysis of the expression state of the linear separable pattern, which main idea is to establish a hyperplane as the decision hyperplane, the decision hyperplane can not only classify all the training samples correctly, but also among training samples make the distance between the points closest to the classification face and the classification face is the maximum. The detailed data can see references [4-5]. Flaw classification task involves training set and test set composed by data examples. Each example in training set includes a target value (class marker) and several attributes (characteristics). The objective of SVM is to produce a target value of data example containing only the attribute among a predicted test set. 4th International Conference on Mechatronics, Materials, Chemistry and Computer Engineering (ICMMCCE 2015) © 2015. The authors Published by Atlantis Press 1055 Supposing that the training sample set is   1 ( , ) n i i i x y  (in which input d i x R  and output   1,1 i y   ), the optimization problem of its hyperplane equation 0 T w x b   can be described as (in which w is the adjustable weight value vector and b is the offset) seeking the minimum of 1 ( ) 2 T w w w   (1) when subjected to the constraint conditions of ( ) 1 1, 2, , T i i y w x b i n     (2) The constrained optimization problem is called original problem, which can be solved by translating it into dual problem utilizing Lagrange multiplier method, namely seeking the maximum of 1 1 1 1 ( ) 2 n n n T i i j i j i j i i j Q y y x x    