Comparison of three methods of 2D defect profile reconstruction from MFL signal
Estimating flaw profiles from measurements is a typical inverse problem in magnetic flux leakage (MFL) testing. Defect profile reconstruction implies the reconstruction of defect parameters and profiles based on detected MFL signals, and it is of importance in achieving the MFL inversion. Through establishing the state-space model of the defect profile and the measured MFL signals, this paper formulates the inverse problem as a tracking problem with state and measurement equations. Three state-space methods, i.e., extended Kalman filter (EKF), unscented Kalman filter (UKF) and particle filter (PF), are employed to solve the inversion problem, which are described as the classical discrete-time tracking problem on the basis of state and measurement equations. The results illuminate that the three state-space approaches are effective and feasible ways of MFL inversion. Furthermore, by comparing the reconstruction performances, it can be found that the particle filter-based inversion approach is superior to the other two methods in actualizing MFL inversion owing to its accuracy and robustness against noise.