Electronique
Les Séparateurs à vaste marge pour la classification des défauts de soudure en radiographie
L’objectif de ce travail porte sur la classification des défauts de soudure dans les images de radiographie des joints soudés en utilisant les caractéristiques géométriques de forme. Ces attributs sont invariants par rapport aux transformations géométriques usuelles de rotation, translation et changement d’échelle. Ils servent à alimenter le classifieur choisi basé sur les séparateurs à vaste marge (SVM). L’algorithme de classification a été appliqué sur une base de données contenant des centaines d’images radiographiques des défauts de soudure qui présentent quatre types de défauts (fissure, manque de pénétration, porosité, et inclusion de solide). Les résultats expérimentaux obtenus confirment l’efficacité des SVMs en termes de précision et temps d’exécution comparé aux réseaux de neurones artificiels Voir les détails
Mots clés : Classification des défauts de soudure dans les images radiographiques, Séparateurs à vaste marge, approche un contre tous
Méthodes de filtrage interférométrique : Etude comparative
Cet article se focalise sur le filtrage du bruit de phase interférométrique, Plusieurs techniques ont été proposées dans ce contexte. Les plus simples sont : Médian et Moyen, ils permettent de filtrer les zones de variations brusques et isolées quant au filtre vecteur qui est un filtre de lissage dont sa forme est adaptée à la variation des phases dans un interférogramme, il convertit la phase en deux fonctions périodiques continues pour régler le problème de discontinuités. Une caractéristique très importante de ce filtre est la préservation des franges avec réduction de bruit. Les méthodes basées sur la décomposition en sous bande ont été développées, leurs principe est de séparer les basses fréquences des hautes fréquences en utilisant les ondelettes. Récemment, de nombreuses études sont concentrées sur les méthodes de filtrage basées sur la décomposition en sous espace. Elles ont l'avantage d'être robustes aux erreurs de recalage, elles emploient la projection de sous-espace signal sur le sous-espace bruit ; qui sont obtenus par la décomposition spectrale de la matrice de covariance, pour estimer la valeur interférométrique optimale de phase. Dans cet article, nous sommes focalisés sur ces méthodes de filtrage dont nous avons proposé une méthode de sélection de sous espace optimal afin de conserver l’information physique ; particulièrement au niveau des zones de transitions des franges. Son principe est inspiré des méthodes de séparation des sources en introduisant l’information de la corrélation entre les pixels des images et la décomposition spectrale Voir les détails
Mots clés : Filrage du bruit de phase, filtre médian, Image Radar SAR, discontinuités de phase
Etude des caractéristiques électriques de cellule solaire à hétérojonction CdS/CIGS (cadmium- sulfide/ cuivre- indium- gallium- sélénium)
Résumé Dans ce travail nous avons effectué l’étude des caractéristiques électriques, une analyse numérique à deux dimensions en utilisant la technologie informatique Aided Design (TCAD Silvaco) logiciel pour CdS/CuInGaSe2 (CIGS) mince cellules solaires à couches est présentée. La structure est composée d'une mince CIGS cellules solaires dans la configuration : ZnO (200 nm) / CdS (50 nm) / CIGS (350 nm) / Mo. ZnO est utilisé pour l'oxyde conducteur (TCO) contacté transparent avant de la cellule. Pour le molybdène de contact arrière (Mo) est utilisé. La couche de la fenêtre de la CdS et la forme de l'absorbeur CIGS le (n-p) semi-conducteurs hétérojonction. La performance de la cellule est évaluée en appliquant les défauts créés dans les joints de grains des CdS poly cristallin et de matériel de CIGS et les CdS d'interface CIGS dans le modèle et les paramètres physiques utilisés dans les simulations TCAD sont calibrés pour reproduire des données expérimentales. Les caractéristiques JV et simulées sous AM 1,5 illumination. Le rendement de conversion (ŋ), de 20,10% est atteinte et les autres paramètres caractéristiques sont simulés : la tension en circuit ouvert (Voc) est de 0,68 V, la densité de courant de circuit (Jsc) est égal à 36.91 mA/cm2 et le facteur de forme (FF) est de 0,80. Les résultats de simulation ont montré que la fraction molaire x de la couche en CIGS a une valeur optimale autour de 0.31 correspondante à une énergie du gap de 1.16 eV, ce résultat est en très bon accord avec celui trouvé expérimentalement. Voir les détails
Mots clés : La modélisation informatique, TCAD Silvaco, paramètres cellulaires des cellules solaires CIGS.
Adaptive algorithms for target tracking.
The quality of the tracking is greatly enhanced by arobust motion estimation.The objective is to develop a target tracking algorithm of amoving object, especially motion estimation of these. To realizethe estimate, we chose stochastic filtering techniques. He concernthe Kalman filter in the linear Gaussian, and sequential MonteCarlo methods in the nonlinear.Representation of state is permanently adapted according oncurrent observations, to best represent the system dynamics. Acomparison of the results given by the extended Kalman filterand particle filter is realized into simulation of the nonlinearsystems target tracking. Voir les détails
Mots clés : stochastic filtring, Kalman filter, Extended Kalman Filter, Monte Carlo Method, Particle filtre, Target tracking
Hybrid Shape Descriptors for an Improved Weld Defect Retrieval in Radiographic Testing
In this paper, four region-based shape descriptors well reported in the literature are used to characterize weld defect types of crack, lack of penetration, porosity and solid inclusion, usually encountered in radiographic testing of welds. The rectangularity and the roundness in the geometric descriptor (GEO) are used in order to propose an hybridization algorithm so that the hybrid descriptor issued from GEO and each of the other descriptors becomes more discriminant in such application where, due to bad radiographic image quality and weld defect typology, the human film interpretation is often inconsistent and labor intensive. According to the results given in the experiments, the efficiency of the proposed hybrid descriptors is confirmed on the weld defects mentioned above where, the retrieval scores are significantly improved compared to the original descriptors used separately. Voir les détails
Mots clés : Radiographic film, weld defects, shape descriptors, hybridization algorithm
A Quantitative Comparative Study of Back Projection, Filtered Back Projection, Gradient and Bayesian Reconstruction Algorithms in Computed Tomography (CT)
Images of the inside of the human body can be obtained noninvasively using tomographic acquisition and processing techniques. In particular, these techniques are commonly used to obtain X-ray images of the human body. The reconstructed images are obtained given a set of their projections, acquired using reconstruction techniques. A general overview of analytical and iterative methods of reconstruction in computed tomography (CT) is presented in this paper, with a special focus on Back Projection (BP), Filter Back Projection (FBP), Gradient and Bayesian maximum a posteriori (MAP) algorithms. Projections (parallel beam type) for the image reconstruction are calculated analytically by defining two phantoms: Shepp-Logan phantom head model and the standard medical image of abdomen with coverage angle ranging from 0 to ± 180° with rotational increment of 10°. The original images are grayscale images of size 128 × 128, 256 × 256, respectively. The simulated results are compared using quality measurement parameters for various test cases and conclusion is achieved. Through these simulated results, we have demonstrated that the Bayesian (MAP) approach provides the best image quality and appears to be efficient in terms of error reduction. Voir les détails
Mots clés : Computed tomography, Bayesian approach, Reconstruction techniques, Filter Back Projection (FBP), Gradient algorithm
Similarity Transformation Parameters Recovery based on Radon Transform. Application in Image Registration and Object Recognition
The Radon transform, since its introduction in the beginning of the last century, has been studied deeply and applied by researchers in a great number of applications, especially in the biomedical imaging fields. By using the Radon transform properties, the issue is to recover the transformation parameters regarding the rotation, scaling and translation, by handling only the image projections assuming no access to the spatial domain of the image. This paper proposes an algorithm using an extended version of the Radon transform to recover such parameters relating to two unknown images, directly from their projection data. Especially, our approach deals with the problem of the estimation accuracy of the rotation angle and its finding in one step instead of two steps as it is reported in the literature. This method may be applied in image registration as well in object recognition. The results are, for the first time, exploited in object recognition where comparison with powerful descriptors shows the outstanding performance of the proposed paradigm. Moreover, the influence of additive noise on registration and recognition experiments is discussed and shows the efficiency of the method to reduce the effect of the noise. Voir les détails
Mots clés : radon transform, 2π-Based Radon transform, Rotation, Scaling and translation transforms, Parameters recovery algorithm, Additive image noise
A comparison of 1D analytical model and 3D Finite Element Analysis with experiments for a Rosen-type piezoelectric transformer
This article is dedicated to the study of Piezoelectric Transformers (PTs), which offer promising solutions to the increasing need for integrated power electronics modules within autonomous systems. The advantages offered by such transformers include: immunity to electromagnetic disturbances; ease of miniaturisation for example, using conventional micro fabrication processes; and enhanced performance in terms of voltage gain and power efficiency. Central to the adequate description of such transformers is the need for complex analytical modeling tools, especially if one is attempting to include combined con- tributions due to (i) mechanical phenomena owing to the different propagation modes which differ at the primary and secondary sides of the PT; and (ii) electrical phenomena such as the voltage gain and power efficiency, which depend on the electrical load. The present work demonstrates an original one-dimensional (1D) analytical model, dedicated to a Rosen-type PT and simulation results are successively compared against that of a three-dimensional (3D) Finite Element Analysis (COMSOL Multiphysics software) and experimental results. The Rosen-type PT studied here is based on a single layer soft PZT (P191) with corresponding dimensions 18 mm x 3 mm x 1.5 mm, which operated at the second harmonic of 176 kHz. Detailed simulational and experimental results show that the presented 1D model predicts experimental measurements to within less than 10% error of the voltage gain at the second and third resonance frequency modes. Adjustment of the analytical model parameters is found to decrease errors relative to experimental volt- age gain to within 1%, whilst a 2.5% error on the output admittance magnitude at the second resonance mode were obtained. Relying on the unique assumption of one-dimensionality, the present analytical model appears as a useful tool for Rosen-type PT design and behavior understanding. Voir les détails
Mots clés : Piezoelectric transformer, Rosen-type transformer, 1D analytical modelling, 3D FEA simulations, Experimental characterisation
Inferential Sensor - Based Adaptive Principal Components Analysis of Mould BathLevel for Breakout Defect Detection and Evaluation in Continuous Casting
This paper is concerned with a method for breakout defect detection and evaluation in a continuouscasting process. This method uses Adaptive Principal Component Analysis (APCA) as a predictor of inputs -outputs model, which are defined by the mould bath level and casting speed. The main difficulties that causebreakout in continuous casting are, generally, phenomenon related to the non-linear and unsteady state of themetal solidification process. PCA is a modeling method based on linear projection of the principal components;the adaptive version developed in this work uses the sliding window technique for the estimation of the modelparameters. This recursive form updates the new model parameters; it gives a reliable and accurate prediction.Simulation results compare PCA, APCA, nonlinear system identification using neural network (NN) and supportvector regression (SVR) methods showing that the APCA gives the best Mean Squared Error (MSE). Based onthe MSE, the proposed approach is analyzed, tested and improved to give an accurate breakout detection andevaluation system. Voir les détails
Mots clés : Soft sensor, continuous casting, Adaptive principal component analysis, breakout detection and evaluation.
Uncertainty estimation of mechanical testing properties using sensitivity analysis and stochastic modelling
This paper is concerned with a method for uncertainty evaluation of mechanical propertiesin metal testing. This method uses a combined approach based on Monte Carlo simulationand Markov Chain (MCMC) as a computing procedure of different uncertainties of mechanicaland metallurgical parameters such as stress, and elongation. The MCMC is a stochasticmethod that computes the statistical properties of the considered states such as the probabilitydistribution function (PDF) according to the initial state and the target distributionusing Metropolis-Hasting (MH) algorithm. Conventional approach is based on the Guide ofUncertainty Measurement (GUM), the uncertainty budget is established for the stress andelongation parameters respectively. A comparative study between the conventional procedureand the proposed method is given. This kind of approaches is applied for constructingan accurate computing procedure of uncertainty measurement of mechanical and metallurgicalparameters. Voir les détails
Mots clés : Markov Chain Monte Carlo (MCMC) Metropolis-Hasting (MH) algorithm Mechanical and metallurgical testing Stress, elongation and hardness measurement Guide of Uncertainty Measurement (GUM)