A Feature Extraction Method of Pipeline Magnetic Flux Leakage Signal based on Expert Experience

Abstract

Considering the complex relationship between the original magnetic flux leakage (MFL) signal collected by the detector in the pipeline and the defect size in the industrial environment, in order to improve the accuracy of defect size estimation, a feature extraction method of MFL signal based on expert experience is proposed. First, a three-axis MFL signal preprocessing method is designed to reduce the abnormal fluctuation and noise interference. Second, based on the principle of MFL signal and domain knowledge, a multi view feature extraction algorithm is proposed, including two parts: static features and dynamic features. Finally, in order to screen out the feature quantity with stronger correlation with defect size, a feature combination method is constructed. The experimental verification is carried out by using the actually collected triaxial MFL signal and the experimental results verify the effectiveness of the proposed method.