Magnetic flux leakage detection in non destructive tests performed on ferromagnetic pieces, using signal processing techniques and data mining

In this paper we propose the use of Support Vector Machine classifiers (SVM) and linear discriminant analysis (LDA) to determine the existence of magnetic flux leakage (MFL) in non-destructive testing (NDT for its acronym in English) performed on ferromagnetic sheets. These signals were provided by the Corporation for Research in Corrosion (CIC) and were acquired on a dyno. The signals are preprocessed to; filter data (ie Wavelet Transform), remove the existing noise (ie thresholding), baseline correction (ie Least Squares Theorem (LST)) and normalize the data (ie First Normal Form). Within the aims of the project are design suitable classifier for each technical proposed for this phenomenon, and a comparison between them to determine which had the best performance.