On magnetic pulse welding machines: deformation measurements and defect detection
Issyan Tekaya  1, *@  , Robert Bouzerar  1@  , Valéry Bourny  2@  , Clément Rabette  1@  , Matthieu Morel  3@  , Olivier Durand-Drouhin  1@  
1 : Laboratoire de Physique de la Matière Condensée  (LPMC)
Université de Picardie Jules Verne : EA2081
33 rue Saint Leu, 80039 Amiens Cedex -  France
2 : Laboratoire des Technologies Innovantes  (LTI)  -  Site web
Université de Picardie Jules Verne : EA3899
33 rue St-Leu, 80039 Amiens Cedex -  France
3 : Plate-forme technologique Innovaltech  (PFT Innovaltech)  -  Site web
Lycée Condorcet
Rond-point Frédéric Joliot Curie 02100 Saint Quentin -  France
* : Auteur correspondant

Magnetic pulse welding is a clean, efficient and extremely quick process that allows the welding of dissimilar metals without the need for additional materials. It basically uses a high pulsed magnetic field to generate eddy currents onto an external piece, that will in turn produce a magnetic pressure through Lorentz forces intense enough to accelerate this piece onto the other, inner piece with immense collision velocities and instantly generate a weld. This work aims to present methods designed both to characterize magnetic pulse machines in order to better understand their functioning but also identify their defects, which could prove detrimental to the quality of the welded pieces, and to measure local deformations using electrical techniques during the welding process.

The first part deals with the electrodynamical characterization of such machines using impedance spectroscopy. Modelling the whole device as an RLC circuit, the values of the three effective parameters R L and C may be deduced from the measured current curves (the current flowing within the coil), which strongly resemble damped sines, reminiscent of the output of a traditional RLC circuit. Comparison may then be held between these values and the ones measured by the use of an impedance analyzer. These parameters are useful as they indicate how the machines should ideally behave and any discrepancy may show up as a defect in their functioning.

A distinct aspect of our work deals with another application of impedance spectroscopy, which as we show, can be used to measure in situ mechanical deformations of coaxial tubes during their welding. Indeed, the tool of choice regarding live deformations measurements usually lies in the Photonic Doppler Velocimetry, a very accurate yet costly optical method. We propose an alternative method for this kind of analysis, based on electrical measurements. Employing the Lagrangian formalism on the physical system composed of two such tubes, the dynamical equations governing the evolution of its impedance of these tubes were established. They allow to couple the electrodynamical properties of the system (its impedance) with its mechanical properties (its deformation). By measuring the impedance variation throughout the process with an impedance analyzer and then numerically solve the boundary value problem (the set of equations), one can obtain the deformation as a function of time. With the theoretical background set, the experimental tests yet have to be carried out and the accuracy and reliability have to be compared to the existing techniques, most notably the PDV.

In the last part, we expose a numerical diagnosis tool useful for the detection and identification of defects of these machines. Indeed, we designed numerical codes written on Mathworks Matlab that process experimental curves (mainly the aforementioned current flowing within the coil) and display diagrams that evidence two kinds of common defects, which are the desynchronization of the power modules and an ineffective distribution of power. Both of these defects lead to losses of energy and therefore a lower magnetic pressure generated. They may therefore greatly lessen the quality of the welding, because the required yield strength may not be attained, for example. Time-frequency analysis tools are employed, mainly the Short Time Fourier Transform and the Wavelet Transform. Different in nature, these tools allow for a convenient representation of the signals with respect to their time evolution and frequency components, simultaneously. We also show how the Wavelet formalism may be applied to detect brutal changes in our measurements.


Personnes connectées : 2