2D-3D Transformations Models for NDT Systems Studies AREZKI DICHE 1*, HASSANE MOHELLEBI 1**, MOULOUD FÉLIACHI 2 1 Département d’Electrotechnique Université de Tizi-Ouzou, B.P. 17 RP, 15000, ALGERIA 2 IREENA(LRTI.-IUT), CRTT, BP 406, 44602 Saint-Nazaire Cedex, FRANCE Abstract: - The present work deals with the exploitation of conformal mapping method principles using established transformations models for eddy current problems in non-destructive testing (NDT) systems. The interest of such a transformation is the generation of electromagnetic field for 2D or 3D axisymmetric problems from only the field distribution known in simpler 2D(x-y) or 3D(x-y-z) geometry. The transformation technique is applied to both geometry and electromagnetic field distribution and validated by comparing the predicted results to reference data for 2D study case [4]. Then the developed models are considered in their 3D forms and applied successfully to 3D structure and quantities. Key-words: Defect analysis; Transformations models; Conformal mapping; Finite element method; NDT. 1 Introduction Transformation methods are a very powerful tool in finite element analysis. An adequate mapping transforms the problem into an easier one or allows to take advantage of the symmetries [1-3]. In this paper transformation models developed to predict electromagnetic behaviour of axisymmetrical 2D or 3D non-destructive testing problems from the knowledge of 2D(x-y) field distribution are presented at first. The procedure resolution converts, at first, the real cylindrical geometry to a plate by means of a geometrical transformation ; in a second step, the electromagnetic problem is solved in terms of vector potential in the plate using finite element analysis ; finally, one transforms the vector potential to obtain its equivalent distribution within the cylindrical device. The predicted results are compared in the 2D case to those given by reference data [4] to obtain the validity of the proposed transformations; after that they were extended and applied to 3D axisymmetric structure and quantities given in the team workshop N° 13. 2 Transformations Models 2.1 Geometrical Transformation Model The geometrical model is obtained with the use of transformation function technique [2]. For its construction, we have to solve the inverse problem, which is consisting on the transformation of the plate into a tube free of deformation with similar width than the initial plate. The method based on the use of the punctual transformations is presented in detail in several references as mathematical and electromagnetic non-destructive evaluation domain [2].The punctual transformation in descriptive geometry is applied to transforms, as a demonstration of the interest method, a line A-B of L dimension as shown following (Fig. 1). Y Y’ M’(x’,y’) R0 A O M(x,y) θ B L X x ’ X’ Fig.1 Transformation of line A-B A point M(x, y) in the A-B segment is transformed into a point M’(x’,y’) in x’,y’, and z’ co-ordinates (Fig.1). The relationship between the initial and the transformed co-ordinates is given by: x' R0 . cos( x ) ; y' R0 . sin( x ) ; R0 R0 with: z' z (1) x R0 Where R0 is the circle radius, x the arc length and knowing that the z component is set invariable. The geometrical model which transforms a plate having some depth into a cylinder is obtained seemly by considering the plate with depth as an infinite superposition of plates without depth. Expressions (2) and (3) give the direct and inverse models respectively. y' x R0 arctan x' P: 2 2 y R0 x' y' z z' x x' ( R0 y ) cos R 0 P-1: x y ' ( R y ) sin 0 R 0 z ' z (2) (3) In formula (2) and (3), x’, y’, and z’ are co-ordinates of a point in the cylindrical geometry and x, y, and z are co-ordinates of the corresponding point in the plate. R0 is the exterior radius of the cylindrical object (Fig. 2). X Y’ -1 P X’ Y Z’ Z P a) b) Fig.2 Position of the problem. 2.2 Victorial Transformation Model The use of the differential geometry theory is unable necessary in searching the vectors transformation. This is consisting on the determination of several derivatives of the transformed components x’, y’ and z’ along the three directions i , j , k as expressed below [3]: ' y' i ' x i x x x' y' j' i y y z' j k x z' j k y ' y' z' k' x i j k z z z Unitarians vectors transformation obtained are then used to deduce a given transformed vector A' , which could be in our case a magnetic vector potential having only one component in the x direction, generally defined by its three components: A' A' x i ' A' y j' A' z k' (4) As the unknown vector studied in our problem is reduced to having only one component in the x direction, the derivative of expression (3) yields to the inverse vectorial model given by formula (6). R y' 0 i x' 2 y' 2 V: R0 y' j 2 x' y' 2 k k x' x' 2 y' 2 .i .j x' 2 y' 2 y' x' 2 y' 2 .i .j x' 2 y' 2 x x i sin R i cos R j 0 0 x x V-1 : j cos i sin j R0 R0 k k (5) (6) For the inverse model we obtain seemly the transformed Unitarians i , j , k vectors as follows: y i x i ' x' x' y j x i ' y' y' ; z j' k' x' z j' k' y' y z k x i ' j' k' z' z' z' Direct model of a given vector transformation A from i , j , k to i , j , k base is obtained by: A Ax i A y j Az k (7) In formula (5) is represented the direct model when considering only the non zero A x component and the derivatives of relation (2). 3 Numerical Analysis The non-destructive testing problem considered in the present work concerns the evaluation of the transducer’s impedance. This necessitates solving the electromagnetic equation with harmonic hypothesis. The numerical approach is given by the time harmonic problem expressed in terms of vector potential: ( A) j 0 . A 0 J (8) 0 = 4π.10-7 H/m ; σ : Electric conductivity. J: current density; 2f , where f is the frequency. Expression (8) is solved using finite element method. The finite element formulation was given as follow: a) b) Fig.3 Cartesian equivalent problem and constructed problem with external defect Fig. 4 shows the vector potential distribution, once it is obtained in Cartesian case (a), its distribution in the tube is obtained just by applying the proposed transformations models (b). A comparison between the predicted distribution and the actual one obtained with finite element analysis (c) show a good agreement. i A A j Adxdy x x 0 y y (6) 0 J dxdy i A (0,0, Az ) and J (0,0, J z ) z Az and J z are components in z direction of magnetic potential and current density vectors. The sensor impedance ( Z ) is then computed using the above impedance’s relationship: a. Cartesian Z jN 2 J S2 2rAds b. Transformed (9) s N and J are total number of turns coil and current density respectively; S: section of the inductor coil; 4 Applications and Results 4.1 2D Case The application considered in this work, also studied in [4], is a non-destructive analysis of a tube by a differential transducer with axisymmetric external defect. The tube parameters are inner radius 9.75 mm, 1.27 mm width, 40 mm high and electric conductivity 1E+6 (S/m). The sensor excited by a current of 5 mA at 100 kHz frequency, is characterised by 7.75 mm inner radius, 0.75 mm width and 2 mm high. Axisymmetric transformed tube Sensor Fig. 4: Vector potential distribution In fig.5 is presented a comparison between Lissajous curves (variation of imaginary part with the real one of the impedance) obtained with the proposed models, posed models, when applying a current density J to the inductor coils, and reference data [4]. -J +J +J 1 1 0. 0. 8 8 plate External defect 0. 0. 6 6 External defect +J -J Im (Z) 0. 0. 4 4 Im(Z) -J c. Axisymmetric 0. 0. 2 2 0 - 0. 0. 2 2 - 0. 0. 4 4 0 Analytical Models Given Results Fig.7 Constructed cylinder Like it is shown in Fig.8.1 and Fig.8.2 which represent the top view and the lateral one of the constructed cylinder, the last is free of deformation. Fig. 5 Lissajous curves for an external defect of 40%. The comparison in terms of real part of the vector potential obtained when using the proposed models and those obtained with finite element analysis along the inner radius of the tube show a good agreement between the results (Fig. 6). Fig.8.1 Top view 2 Transformations models f.e.m 1.5 real part of the vector potential [ T.m] Fig.8.2 Lateral view x10-8 1 0.5 Applying the direct model to the constructed cylinder leads to the original plan configuration, i.e. a plate scanned with a differential sensor, as it is represented in Fig.9. 0 -0.5 -1 -1.5 -2 0.02 0.025 0.03 0.035 0.04 z [m] 0.045 0.05 0.055 0.06 Fig.6 Comparison between real parts of vector potential 4.2 3D Case 4.2.1 Geometrical Model The upcoming application is considered in order to extend the validity of the developed models to the 3D cases. In Fig.8 is shown the constructed cylinder with the differential sensor. It is obtained using the geometrical inverse model to transform the plate and the differential sensor defined above. Fig.9 Original configuration This allows us to conclude the validity of the proposed geometrical model. 4.2.2 Vectorial Model In this case the inverse vectorial model is applied to the results given in team workshop N°13 for the magnetic flux density. The geometric characteristics of the system are as follow. z Steel chanel x y y Plate Coil 2 Cylindrical object Given results for the plate Magnetic induction [T] 1.8 1.6 a) Front view b) Top view 1.4 1.2 Fig.10 Geometrical arrangement 1 The given results are depicted in Table 1. Coordinates [mm] N° 1 2 3 4 5 6 7 X Y -25 <= x <= 25 0 <= y <= 1.6 0.8 0.6 0 Induction in [T] Z 0 10 20 30 40 50 60 1.333 1.329 1.286 1.225 1.129 0.985 0.655 The obtained results are given in Fig.11 below. One can see that a good repartition of the magnetic flux density surrounding a cylinder object. 0.01 0.02 0.03 High [m] 0.04 0.05 0.06 Fig.12 Predicted results for the cylindrical object In a second step, the direct model is applied to the obtained results when using the inverse vectorial model. The planar repartition of the magnetic flux density is depicted in Fig.13. This shows a good repartition of the vectors. 0.07 0.06 0.05 Z 0.04 0.07 0.03 0.06 0.05 Z 0.02 0.04 0.01 0.03 0.02 0 5*10-4 0.01 0 0 0.01 -5 0.005 0.01 0.005 0 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 Y X 0 -0.005 Y -0.005 -0.01 X Fig.11 Predicted repartition of the magnetic flux density in the cylindrical object The result of the magnetic flux density obtained using the inverse vectorial model is represented in Fig.12, where some difference could be note between the initial and the transformed quantities. Fig.13 Planar repartition of the magnetic flux density The bijectivity of the proposed models is checked when using the inverse model, for the predicted results in the cylindrical body. We note that the initial values in the plate are the same than the results obtained when the inverse models are applied to the geometry and flux density values, like it is represented in Fig.14. This offers some interesting perspectives 3D numerical solutions in NDT systems or Non Destructive Evaluation NDE) 1.4 1.3 Magnetic Induction [T] 1.2 1.1 1 Analytical Models [T] Given results 0.9 0.8 0.7 0.6 0 0.01 0.02 0.03 High [m] 0.04 0.05 0.06 Fig.14 Bijectivity of the proposed models 5 Conclusion Transformation functions which allow to predict the electromagnetic phenomena behaviour in a cylindrical object with external axisymmetric defect by means of equivalent plate geometry, in nondestructive testing domain, are presented. The validity of results is obtained by comparison with reference data, in 2D case. Then the proposed models were extended to 3D structure and applied successfully to magnetic flux density quantities (3D). References [1] A. Diche, H. Mohellebi, M. Féliachi and G. Berthiau, Development of Cartesianaxisymmetric Transformations Models for NonDestructive Testing Problems, ISEM’03, Versailles, France, 12-14 Mai 2003, 144-145. [2] S. Mziou, Application de transformations mathématiques à la résolution des équations de Maxwell pour le contrôle des tubes de générateur de vapeur par une sonde ponctuelle, Stage de DEA Université Paris-Centre Jussieu Août 1999, LCME, CEA Saclay. [3] F. Henrotte, B. Meys, H. Hedia, P. Dular and W. Legros, Finite Element Modelling with Transformation Techniques, IEEE Transaction on Magnetics, vol.35, No.3, May 1999, pp 1434-1437. [4] J. L. Thomas, Modélisation Simplifiée du contrôle par Courants de Foucault de Tubes de Générateur de Vapeur, Rapport de Stage, ESA IGELEC, Université de Nantes, France, 1998.
© Copyright 2026 Paperzz