Wavelet transform and neural network based 3D defect characterization using magnetic flux leakage

Abstract

Magnetic flux leakage (MFL) technique is commonly used for inspection of gas transmission pipelines. MFL signal is used to identify and characterize defects in the pipeline by estimating their length, width and depth (LWD). Knowledge of LWD alone is highly inaccurate and coarse compared to actual 3D geometry of the defect for predicting the maximum allowable operating pressure (MAOP) of the pipe. However, the inverse problem associated with prediction of 3D geometry is not only ill conditioned, but also involves complex numerical computation. As a result, little research has been done in this area. Author has published two different methods of dealing with this problem in collaboration with fellow researchers. This paper reviews the two approaches for estimating 3D depth profile of a defect from the corresponding MFL signal based on radial basis function neural network (RBFNN) and discrete wavelet transform (DWT).