Electronique
Deconvolution of closer ultrasonic backscattered echo signals from composite materials
In this paper, we present an approach of deconvolution problem from the estimation of superimposed signals based on the Expectation Maximization Algorithm (EM algorithm) and Maximum Likelihood Estimation. The idea is to decompose the observed data into their signal components and then to estimate separately the parameters of each component signal. This algorithm iterates back and forth using the current estimated parameter in order to decompose the observed data and thus to increase the likelihood of the next estimated parameters. We initially apply this method to simulated signals with additional structural noise. These signals contain several echo defects, closer between them. This stage permits to see the robustness of the developed algorithm. Thereafter, we validate all simulated results by experimental results obtained on composite material with and without delamination defects. Voir les détails
Mots clés : Ultrasonic NDE, composite materials, Deconvolution, EM algorithm
Texture analysis for flaws detection in ultrasonic images
In this paper, we present two approaches to detecting a flaw in TOFD (Time Of Flight Diffraction) type ultrasonic image based on texture features. Texture is one of the most important features used in recognizing patterns in an image. The paper describes texture features by two methods: Multiresolution analysis such as wavelet transforms and Gabor filters bank. These filters are based on Gaussian shaped band-pass filters, with dyadic treatment of the radial spatial frequency range and multiple orientations. These filters represent an appropriate choice for tasks requiring simultaneous measurement in both space and frequency domains. The most relevant features are used as input data on a Fuzzy C-Mean clustering classifier. We use two classes: 'defects' or 'no defects'. The proposedapproach is tested on the TOFD image achieved at laboratory and industrial field. Voir les détails
Mots clés : Texture analysis, NDE, TOFD image, Wavelet transform, Gabor filters, Fuzzy logic
Location of material flaws using Split Spectrum Processing and wavelet transform
The Non Destructive Testing by ultrasounds is based on the detection and theinterpretation of ultrasonic wave reflected by material flaws. Various signal processing techniques were introduced. They are based on frequential analysis in order to increase the defects detection and to improve theirs localization. It is quite allowed today that the signalrepresentations, jointly in time and frequency, are of interest. They give a natural description of the nonstationary signals whose frequency varies in time or comes from transient states. In this work, we develop echoes detection technique drowned in material structural noise which can induce in error, the controller experts in the results interpretation. So, we show that theapplication of algorithms based on Split Spectrum Processing (SSP) and the Discrete Wavelet transform (DWT), can increase detection possibilities. A comparative study is carried out between SSP with Q constant and DWT in particular Symlets wavelets, Daubechies wavelets and Coiflets wavelets with several orders associated to some thresholding algorithms. Allthese methods are developed and applied to ultrasonic signals with additional modelled structure noise. The simulation results are validated by experimental results obtained on steel material. Voir les détails
Mots clés : Denoising, Discrete wavelet transform, Signal processing, Ultrasonic
Estimation des signaux d’échos ultrasonores rapprochés
Cet article présente un modèle d’estimation des échos ultrasonores pour l'évaluation des signaux superposés basés sur l'algorithme EM et le maximum de vraisemblance. L'idée est de décomposer les données observées en leurs composants de signal et puis d'estimer les paramètres de chaque composant de signal séparément. L'algorithme réitère dans les deux sens, en utilisant les estimations courantes de paramètre pour décomposer les données observées mieux et pour augmenter ainsi la probabilité des prochaines évaluations de paramètre. Nous appliquons la méthode à des acquisitions des signaux sur une pièce d’acier avec des défauts rapprocher Voir les détails
Mots clés : Estimation, ultrason, maximum de vraisemblance, algorithme EM
Deconvolution of ultrasonic echoes using bernoulli-gaussian processes for composite materials inspection
In this work, we present an approach of deconvolution ill-posed problems of superimposed signals in time. A priori information must be taken into account to solve this problem. The a priori information translates the physical properties of the ultrasonic signals. The defect impulse response is modelled as a Bernoulli-Gaussian sequence. Deconvolution becomes the problem of detection of the optimal Bernoulli sequence and estimation of the associated complex amplitudes. We initially apply this method in order to simulate signals with additional structural noise. These signals contain several echo defects, closer between them. This stage permits to evaluate the robustness of the developed algorithm. Thereafter, we validate all simulated results by experimental results obtained on composite material with and without delamination defects. Voir les détails
Mots clés : Ultrasonic NDE, composite materials, blind deconvolution, processes BG
Comparison of Signal Processing Techniques for Ultrasonic Inspection of Composite Materials
In this paper, two methods permitting the detection and estimation of delamination defect echoes are developped. These methods are based on signal processing techniques such as non linear filtering realised by Split Spectrum Processing (SSP) and estimation by EM algorithm. A simulation study of defect detection is realised and results are validated by experimental results obtained on two types of composite taken from aircraft like fiber-reinforced multi-layered composite materials. Voir les détails
Mots clés : Ultrasonic NDE, composite materials, SSP, EM algorithm
Etude comparative des algorithmes EM et SSP dans la détection des échos ultrasonores superposés
Dans ce travail, nous proposons de développer deux approches pour le problème d’échos ultrasonores superposés dans le temps, la première est une méthode de filtrage non linéaire réalisée par la technique Split Spectrum Processing (SSP). La seconde méthode est une technique d’estimation basée sur l'algorithme EM. Les deux approches sont appliquées à des signaux expérimentaux issus du contrôle ultrasonore sur une pièce d’acier avec des échos de défauts rapprochés. Voir les détails
Mots clés : ultrasons, Filtrage non linéaire, SSP, Estimation, algorithme EM
Radiographic Image Segmentation Based on Gaussian Mixture Model
In this work, we propose to use an image segmentation method based on Gaussian mixture model. The observed image is considered as a mixture of multivariate densities and the mixture parameters are estimated by the expectation maximization (EM) algorithm. The segmentation is completed by clustering each pixel into a component according to the maximum likelihood (ML) estimation. This method has been applied to a variety of radiographic images of weld defects and satisfactory segmentation results have been reported. Voir les détails
Mots clés : — expectation maximization algorithm, fuzzy C-means algorithm, Gaussian mixture model, image segmentation, radiographic images, weld defect
Unsupervised Algorithm for Radiographic Image Segmentation Based on the Gaussian Mixture Model
In this paper we study an unsupervised algorithm for radiographic image segmentation, based on the Gaussian mixture models (GMMs). Gaussian mixture models constitute a well-known type of probabilistic neural networks. One of their many successful applications is in image segmentation. Mixture model parameters have been trained using the expectation maximization (EM) algorithm. Numerical experiments using radiographic images illustrate the superior performance of EM method in term of segmentation accuracy compared to fuzzy c-means algorithm. Voir les détails
Mots clés : weld defect, radiographic images, image segmentation, Gaussian mixture model, expectation maximization algorithm, fuzzy C-means algorithm
Parametric active contour for boundary estimation of weld defects in radiographic testing
In this paper we present a new approach to deal with the defects contours estimation problem in radiographic images using parametric active contours. In this approach we exploit the performance of the GVF as external force and enhance it by joining to it external adaptive pressure forces which bring speed to the snake evolution and less sensitivity to the snake initialization and provides capability of tracking the concavities. Voir les détails
Mots clés : Active contour, GVF, Pression forces, radiographic images