The principle of Non Destructive Testing (NDT) by using magnetic flux leakage (MFL) consists to magnetize a magnetic component to be inspected by a strong magnetic field and to detect with a magnetic sensor the magnetic flux lines which are leaking from part due to a defect. MFL methods are usually used during the process of manufacture of ferromagnetic pipes by our partner, the Vallourec Group, the leader in the world in manufacturing of pipes. To improve NDT systems in manufacturing plants, the CEA-LIST and the research center of Vallourec are working together to develop fast simulation tools dedicated to virtual testing of ferromagnetic pipes. The main experimental system concerns the detection of longitudinal defect. The modeling problem is to solve an electromagnetism problem with a complex geometry in the magneto-static nonlinear regime. Eddy currents induced by the motion of the pipe with respect to the magnetizing system are neglected.In this context, a semi-analytical approach based on integral equations (IE) has been chosen. The goal of some works which have been carried out in this PHD thesis is to address 3D complex geometries but, in first a step, limited to the linear regime. However, the non-linear behavior of the relationship which links the magnetic flux density and the magnetic field inside a ferromagnetic material must be considered when choosing the 3D formulation of the problem.After a study about influent parameters of the experimental system, carried out by using finite elements computations, we have considered two strategies for modeling. The first one has consisted to build up a strategy which consists to join the 2D numerical model, existing in the laboratory to an extension model from 2D to 3D. This approach was a priori quite simple but the lack of generality of this approach leads us to suggest another strategy which results in solving the complete 3D magneto-static problem. This formulation is based the integral equation formalism implying an auxiliary scalar quantity: the magnetic surface charge density. In order to be able to solve a nonlinear problem in the future, the chosen numerical scheme we have adopted is based on two hypothesis: the ferromagnetic part is firstly divided into a finite number of small hexahedral cells in which the relative magnetic permeability is supposed to be constant and secondly, the unknowns of the problem, the surface charge densities on the facets of each cell are expanded by using basis functions of zero order. Thanks to this limitation, the singular integrals can be analytically computed. Some simulation results confirm the validity of the implemented numerical model. This model presents some limitations at the moment in the cases of a workpiece with a defect but it can provide quite good results without any defect. Several geometries have been addressed by using Gmsh, free meshing software. Moreover, the final numerical model can be included into an iterative process to deal with non-linear cases. The limitations due to the EI formalism can be overcome today by using some compression matrix methods. Presently, this model is a good candidate for virtual NDT for cylindrical and planar geometries by magnetic flux leakage.
