Research On Extracting And Analyzing Features Of Pipeline Magnetic Flux Leakage Data

Pipeline transportation is an extremely important mode of transportation which has been playing a more and more increasingly significant role in the industrial production and the development of the national economy. However, with the time passing by, the inner and outer walls of pipe will occur inevitably a degree of corrosion. Defects generated due to corrosion will cause catastrophic accidents. Therefore, the using of magnetic flux leakage method to detect defects has great significance.Accurately estimating the shape and location of defects by test data is the core of magnetic flux leakages. Therefore, distinguish abnormal data and extracting the features of abnormal data are the most key links in the entire process of magnetic flux leakage test. For this important link, this paper propose a method of distinguish abnormal data extracting the features of abnormal data and analyzing the features. The following three issues have been studied:distinguish abnormal data of single defect and extracting the features, the processing method of the Magnetic flux leakage testing data of detects located in the complex position and analyzing the correlation of abnormal data’s features and the shape of defectsFirstly, the value differential algorithm has been proposed to distinguish abnormal data base on the algorithm of threshold value and the algorithm of differential threshold. Extracting follow features of abnormal data:the peak and valley value, the valley and valley value, the space between two inflection points, the waveform area, the waveform energy and so on. For the most difficult issue that extracting the space of two inflection points, this paper propose a method to calculate the location of inflection points based on wavelet transform.Not only have some isolated defects in the inside and outside the wall of pipes. There are a lot of defects in the weld or in more complex position. In order to solve this issue, the judgment methods of various pipe welds have been proposed and the method which processing test data of defects in the weld has been studied. If two defects too close, abnormal data of the two defects will overlap. For extracting the features, separating the overlapped data of two defects and using data fitting to fill the missing part of separated abnormal data.Finally, intuitive analyze features of abnormal data and obtain the features which change largely with the change of defect shape. Analyzing and processing these features by principal component analysis to decrease dimensionality of these features and obtain a few principal components which can take the place of these features. At the same time, respectively ascertain the features which have significant role in estimating the length width and depth of the defects.