Non-destructive testing (NDT) is the process of non-invasive material examination. Within this field, ultrasonic inspection is one method of examination used to detect flaws in structural and functional industrial components, to assess their structural integrity and fitness for service. Conventional NDT ultrasonic array techniques transmit on multiple elements in parallel, according to a focal law, which facilitates beam steering, focussing and scanning within the test component. Received signals are then 'stacked' to generate images of the test component. With the advent of affordable high speed computing, novel data acquisition techniques based on sequential transmission are now able to be developed, which allow images to be generated using advanced signal processing and image reconstruction algorithms. One such data acquisition technique known as Full Matrix Capture (FMC), has received considerable research attention in recent years, largely because it allows fully focused images of test components to be generated. This project provides an improved understanding of the FMC technique and associated signal processing algorithms. It achieves this through the development of novel inspection techniques and signal processing algorithms. Collectively algorithms developed within this work were termed Sequential Phased Array (SPA). Initially comparisons were made between conventional ultrasonic techniques and the SPA algorithms in terms of image quality and speed of inspection. A novel approach was then suggested to facilitate inspection through dual-layered media, separated by a refractive interface using the SPA algorithms. The use of sparse arrays to enhance the speed of inspection using the SPA algorithms was also investigated, and the trade-off between speed of inspection against image degradation was addressed. Finally a novel approach to calibration of a FMC based system was developed, in order to provide uniform image sensitivity during inspection.

Les techniques de contrôle non-destructif s’appuyant sur des traducteurs ultrasons multi-éléments sont maintenant de plus en plus utilisées dans l’industrie. Ces traducteurs offrent en effet de nombreux avantages, en particulier une polyvalence d’utilisation et une adaptabilité à des configurations complexes. De plus des techniques d’acquisition avancées, telle que la Full Matrix Capture, permettent d’avoir accès à des données beaucoup plus riches et complètes que celles fournies par les techniques conventionnelles, offrant ainsi de nouvelles possibilités en terme de localisation et de caractérisation de défauts. Les travaux de cette thèse ont pour objectif de mettre en œuvre des méthodes d’imagerie et de reconstruction adaptées aux diverses acquisitions multi-éléments possibles. Pour cela, nous proposons différentes méthodes dérivées de l’approche dite de focalisation synthétique. Ces méthodes s’appuient sur des modèles directs développés au CEA LIST et implémentés dans la plateforme CIVA. Ces derniers peuvent, en outre, prendre en compte les configurations de contrôle complexes, notamment en ce qui concerne l’état de surface -irrégulier- des pièces inspectées. Dans un premier temps, nous nous sommes intéressés à la localisation de diffracteur quasi-ponctuel. Sous cette hypothèse, nous avons évalués les performances des méthodes proposées lorsqu’elles sont appliquées à différents types d’acquisitions. La suite des travaux s’est recentrée sur l’étude de la Full Matrix Capture couplée à l’algorithme dit de « Focalisation en Tous Points (FTP) », cette combinaison ayant montré de bonnes performances en terme de localisation. Dans un second temps, nous avons prolongé notre approche à la reconstruction de défauts étendus et à la prise en compte de trajets ultrasonores multiples. En particulier, nous avons appliqué la généralisation de l’algorithme FTP à la localisation et à la caractérisation d’entaille proche du fond de la pièce à l’aide d’échos de coin, échos dus à des réflexions successives de l’onde ultrasonore sur le fond et le défaut. Les différentes méthodes envisagées dans ces travaux ont été appliquées avec des données simulées ainsi qu’avec des mesures expérimentales. / Techniques for nondestructive testing based on ultrasonic phased array probe are now increasingly used in industry. These probes offer many advantages, in particular use versatility and adaptability to complex configurations. More advanced acquisition techniques, such as “Full Matrix Capture”, provide access to collection of data richer and more complete than those provided by conventional techniques, offering so new possibilities in terms of localization and characterization of defects. The work of this thesis aims to implement imaging and reconstruction methods for various operating modes of UT array inspection. For this, we propose different methods derived from the so-called synthetic focusing. These methods rely on forward models developed at CEA LIST and implemented in the CIVA software platform. These models can also take into account the complex inspection configurations, especially in regard to the -irregular- surface inspected parts. At first, we were interested in locating point-like scatterer. Under this hypothesis, we evaluated the performance of the proposed methods when they are applied with different types of inspection operating mode. Next work has been refocused on the study of the Full Matrix Capture coupled to the algorithm called Focalisation en Tous Points (FTP), this combination has shown good performance in terms of defect localization. In a second step, we generalized our approach to extended defects and multiple ultrasonic paths. In particular, we applied the latest FTP algorithm to the localization and characterization of notch near the bottom of the part with corner echoes, echoes due to successive ultrasonic wave reflections on the bottom and defect. The different methods considered in this work have been applied with both simulated data and experimental measurements.

A crescente tendência de utilização de materiais austeníticos soldados e cladeados em componentes críticos em alguns setores industriais, como nas indústrias de óleo&gás e nuclear, leva a um aumento na demanda sobre ensaios não-destrutivos confiáveis na avaliação de sua integridade estrutural. Dentre os métodos utilizados na inspeção de soldas cladeados austeníticos estão os métodos de ultrassom por phased array, que são normalmente utilizados na detecção e localização de defeitos. No entanto, componentes com esse tipo de microestrutura são difíceis de inspecionar por phased array devido a anisotropia e inomogeneidade causadas pela microestrutura de grãos grosseiros que costumam levar ao aumento do nível de ruído, ao deslocamento de indicações e ao surgimento de indicações falsas. Sendo assim, a seleção de um método de phased array apropriado precisa levar em conta a habilidade do método em superar os problemas causados pela anisotropia e inomogeneidade. Esta tese apresenta dois métodos de imagem por phased array ultrassônico não-convencionais pensados como formas de ajudar na determinação da integridade de componentes onde soldas e cladeados austeníticos estão presentes. Ambos os métodos tem como base o método de foco total (TFM), sendo que o primeiro é uma extensão do método de leis de atraso adaptativas chamado Método de Foco Total de Atraso Adaptativo (ADTFM) e o segundo método usa fatores de coerência associado à imagens de TFM. A partir dos métodos de imagem aplicados é possível aumentar significativamente a qualidade das imagens por ultrassom em comparação com as imagens padrão obtidas por TFM, especialmente quando foi possível utilizar ambos os métodos combinados. / The increasing trend to use austenitic welded and cladded materials in critical components employed in some industrial sectors, such as the oil&gas and nuclear industries, leads to an increasing demand for their non-destructive assessment by reliable non-destructive methods. Among the methods used to access the integrity of austenitic welds and claddings are the Ultrasonic Phased Array methods, which are usually used to detect the presence and determine the position of defects. However, austenitic welds and claddings are challenging to inspect with Phased Array methods due to the anisotropy and inhomogeneity caused by their coarse grain microstructure, which is capable of increasing noise levels, misplace indications and create false indications. Therefore, the selection of an appropriate phased array method needs to take into account the method’s ability to overcome the impairment caused by anisotropy and inhomogeneity. This thesis presents two non-conventional methods based on ultrasonic phased array imaging techniques designed to assist the structural integrity assessment of components where austenitic welds and clads are present. Both proposed methods are based on the Total Focusing Method (TFM); the first approach is an expansion of the adaptive delay laws concept named Adaptive Delay Total Focusing Method (ADTFM), while the second method uses the coherence weights combined with the TFM images. From the imaging methods applied it was possible to significantly increase the quality of the ultrasonic images in comparison with the standard TFM, primarily when it was possible to combine both approaches.

Les techniques de contrôle non-destructif s'appuyant sur des traducteurs ultrasons multi-éléments sont maintenant de plus en plus utilisées dans l'industrie. Ces traducteurs offrent en effet de nombreux avantages, en particulier une polyvalence d'utilisation et une adaptabilité à des configurations complexes. De plus des techniques d'acquisition avancées, telle que la Full Matrix Capture, permettent d'avoir accès à des données beaucoup plus riches et complètes que celles fournies par les techniques conventionnelles, offrant ainsi de nouvelles possibilités en terme de localisation et de caractérisation de défauts. Les travaux de cette thèse ont pour objectif de mettre en œuvre des méthodes d'imagerie et de reconstruction adaptées aux diverses acquisitions multi-éléments possibles. Pour cela, nous proposons différentes méthodes dérivées de l'approche dite de focalisation synthétique. Ces méthodes s'appuient sur des modèles directs développés au CEA LIST et implémentés dans la plateforme CIVA. Ces derniers peuvent, en outre, prendre en compte les configurations de contrôle complexes, notamment en ce qui concerne l'état de surface -irrégulier- des pièces inspectées. Dans un premier temps, nous nous sommes intéressés à la localisation de diffracteur quasi-ponctuel. Sous cette hypothèse, nous avons évalués les performances des méthodes proposées lorsqu'elles sont appliquées à différents types d'acquisitions. La suite des travaux s'est recentrée sur l'étude de la Full Matrix Capture couplée à l'algorithme dit de " Focalisation en Tous Points (FTP) ", cette combinaison ayant montré de bonnes performances en terme de localisation. Dans un second temps, nous avons prolongé notre approche à la reconstruction de défauts étendus et à la prise en compte de trajets ultrasonores multiples. En particulier, nous avons appliqué la généralisation de l'algorithme FTP à la localisation et à la caractérisation d'entaille proche du fond de la pièce à l'aide d'échos de coin, échos dus à des réflexions successives de l'onde ultrasonore sur le fond et le défaut. Les différentes méthodes envisagées dans ces travaux ont été appliquées avec des données simulées ainsi qu'avec des mesures expérimentales.

[pt] Sinais ultrassônicos são amplamente utilizados na indústria para detecção de defeitos e caracterização de materiais. Neste âmbito, o método de focalização total (TFM) é comumente aplicado no pós-processamento dos dados ultrassônicos. Estes dados são adquiridos por transdutores multielementos do tipo phased array pela técnica de Captura de Matriz Completa (FMC). Embora a técnica FMC-TFM seja amplamente empregada, a energia da onda transmitida ao material é limitada pela utilização de um único elemento do transdutor, podendo fornecer uma baixa relação sinal-ruído (SNR) na imagem reconstruída. Para superar essa limitação e melhorar a qualidade das imagens TFM, propõese neste trabalho o uso da técnica Fontes Virtuais combinada com dois métodos diferentes de melhoria do SNR: Decomposição do Operador de Reversão Temporal (DORT) e Correlação Espacial de Sub-abertura (SASACI). Neste último, foram ainda propostas alterações para tornar a abordagem original mais robusta. Ambas as propostas de combinação foram aplicadas e avaliadas para melhorar as imagens de caracterização de múltiplos defeitos. Tais propostas foram avaliadas por meio de simulações e experimentos. As simulações consideraram um modelo de elementos finitos de um bloco de aço contendo quatro furos simulando defeitos volumétricos no material. Vários níveis de ruído foram adicionados aos sinais simulados para se avaliar o desempenho das combinações propostas em um ambiente com alto nível de ruído não-correlacionado. As combinações propostas nos dados de FMC simulados mostraram que o uso da técnica Fontes Virtuais combinada com a técnica DORT foi capaz de aumentar o SNR em 21.5 dB, enquanto a combinação de Fontes Virtuais com o SASACI foi capaz de aumentar o SNR em 76.2 dB. Este último resultado está 16.3 dB acima da soma dos ganhos individuais de cada técnica, enquanto o primeiro está 7.4 dB acima. Isso indica um efeito sinérgico no aumento da qualidade da imagem para ambas as combinações. Além disso, o uso das Fontes Virtuais mostrou-se capaz de melhorar a imagem reconstruída, onde inicialmente não é possível distinguir o ruído do sinal de interesse. A avaliação experimental foi realizada sobre o bloco de aço de quatro furos com as mesmas propriedades do bloco simulado. Ao contrário da simulação, os dados coletados não estão sujeitos à altos níveis de ruído não-correlacionado. Portanto, em ambas as combinações, o emprego da técnica Fontes Virtuais não resultaram em ganhos significativamente superiores aos ganhos individuais de cada técnica. Os métodos DORT e SASACI sem Fontes Virtuais aumentaram o SNR em 7.5 dB e 75.0 dB, respectivamente, enquanto que, com o seu uso, esse aumento foi de 9.2 dB e 74.1 dB. Os resultados do SASACI, quando comparados ao DORT, se mostraram evidentemente superiores tanto na simulação quanto na avaliação experimental. Ambos os métodos proporcionaram melhoria da qualidade de imagens TFM e, portanto, promissores para serem aplicados em ensaios não destrutivos. / [en] Ultrasonic signals are widely utilized in the industry for detecting defects and material characterization. In this context, the Total Focusing Method (TFM) is commonly applied for post-processing of ultrasonic data. Those are acquired by phased array transducers through the Full Matrix Capture (FMC) technique. Although the FMC-TFM technique is widely employed, the energy of the wave transmitted to the material is limited by the utilization of a single element of the transducer, which can provide a low Signal-Noise Ratio (SNR) of the reconstructed image. To circumvent this limitation and enhance TFM image quality, this work proposes the use of the Virtual Sources technique combined with two different methods for SNR improvement: Decomposition of the Time Reversal Operator (DORT) and Spatially Averaged Sub-Aperture Correlation Imaging (SASACI). In the latter, we also propose modifications to make the original approach more robust. Both combinations were implemented, applied and evaluated for the enhancement of images for characterization of multiple defects. These proposals were all assessed through simulations and experiments. The simulations considered a finite element model of a steel block containing four holes emulating volumetric defects on the material. Several noise levels were added to the simulated signals aiming for performance assessment in an environment with high levels of non-correlated noise. The proposed combinations in the simulated FMC data showed that the utilization of the Virtual Sources technique combined with DORT increased the SNR up to 21.5 dB, while the combination of Virtual Sources with SASACI was capable of increasing SNR up to 76.2 dB. The second combination is 16.3 dB above the sum of the individual gains for each technique applied separately, while the first combination is 7.4 dB above. This implies a synergistic effect in the improvement of image quality for both proposed combinations. Moreover, the utilization of the Virtual Sources technique was capable of enhancing the reconstructed image, where it is not possible to initially distinguish the noise of the signal of interest. The experimental evaluation was realized on the four-hole steel block with the same properties of the simulated block. Unlike the simulation, the collected data were not subjected to high levels of noncorrelated noise. Thus, for both combinations, the utilization of the Virtual Sources technique did not result in significantly superior gains when compared to the individual gains obtained from each technique individually applied. The use of DORT and SASACI without Virtual Sources caused an SNR increase of 7.5 dB and 75.0 dB, respectively, while this increase was of 9.2 dB and 74.1 dB when the Virtual Sources was used in combination. The results obtained from SASACI, when compared to DORT, were evidently superior for both simulation and experimental evaluation. Both methods showed potential for improvement of TFM image quality, and thus, are promising for application in nondestructive evaluation.