Guided Wave Testing (GWT) is a relatively new development in non-destructive testing. Conventional Ultrasonic Testing (UT) methods are operated at high frequencies (MHz) and are capable of detecting very small (down to micrometre-scale) flaws within a range of millimetres from a transducer. GWT, however, is carried out at lower frequencies (kHz) and is capable of highlighting the position of volumetric structural detail and discontinuities, such as gross corrosion at a minimum of 9% of the cross-sectional area, tens of metres from a test location. Conventional ultrasonic testing relies on the transmission of bulk waves whereas GWT employs so-called ultrasonic guided waves (UGW). To simplify UGW inspections, several tests are conducted sequentially at a range of different excitation frequencies. The frequency bandwidth of each of these tests needs to be controlled to avoid complexities caused by the frequency dependent nature of the propagation of guided waves. This gives rise to the current GWT inspection procedure, where a number of different narrowband tests are conducted at several distinct frequencies. It is also found that different test circumstances (such as pipe coating or defect type) are inspected more easily with certain excitation frequencies than with others - and the optimum frequency can not always be predicted ahead of time. Thus, where time allows it is often beneficial to carry out a frequency sweep, whereby a large range of excitation frequencies are incrementally generated - for example, from 20 to 80kHz in 1kHz steps. This research proposes a novel approach to the existing pipeline inspection procedure by utilising the information contained within a broadband response. The overarching proposition given by this research is that the current collection procedure be entirely rewritten. This thesis will present ideas related to every area of the inspection procedure beginning with the tuning of excitation signals and concluding with recommendations on how tooling and excitation configuration can be modified to further optimise the technique for broadband excitation.

<p>To be able to ensure quality, efficiency and safety in nuclear reactors, non-destructive evaluations (NDE) are performed. The moulding technique, which has been studied in this project, is an NDE method used to verify surface breaking cracks at various objects in reactor vessels.</p><p>The idea of moulding is to receive a copy of the replicated surface for microscopic analysis. Within forensic science the moulding technique is used at crime scenes to collect evidence and tie suspects to crimes. Underwater moulding, however, is a newly developed technique and WesDyne TRC is a pioneer in offering services within moulding for underwater purposes.</p><p>This project was initiated by WesDyne TRC to further their knowledge within the moulding technology. In the project, studies have been made at three important parameters effect on cast quality using three different polymer compounds. Problems during moulding, such as crack detection failures and bubbles in the casts, raise the question whether the underwater moulding technique can be trusted to detect cracks.</p><p>Results from the experiments led to a greater insight into the problem with receiving high quality casts during underwater moulding. Only if a satisfactory cast is made, the moulding method can be trusted to detect defects down to the detection target in both dry and wet environment.</p><p>To increase the surface quality of underwater casts a suggestion for mould design and a recommended moulding method was developed. In addition, one of the polymer compounds approved for use, turned out not to be suitable for underwater moulding.</p>

The heterogeneous and anisotropic nature of wood material creates additional design challenges not present with the use of other structural materials such as steel and aluminum. The natural variation in the physical properties of wood members requires that the specified strengths and resistances used for design calculations be based on the quantities measured for the fifth percentile of all wood materials tested. The result is that design may be unnecessarily conservative and subsequently inefficient. The same properties that cause uncertainty surrounding the physical properties of biological materials also create difficulty in applying non-destructive evaluation techniques. Strain measurement is one particular technique that is extremely valuable for materials of known and consistent stress-strain relationships, but whose usefulness is diminished when applied to biological materials. To demonstrate the need for more accurate strain measurement in light-framed structures, the predictive calculations and structural modelling of a post-framed building was compared to its demonstrated performance. The analysis did not adequately reflect the actual performance of the building, and it was determined that additional monitoring of light-framed buildings through systems such as strain measurement was required to gain a better understanding of the performance characteristics in order to optimize evaluation techniques. This project aimed to develop a system that accurately measures strain in dimensional lumber of different types, which in turn will enable researchers to enhance monitoring the performance of light-frame structures and optimize design analysis and structural modelling techniques. The development of a methodology that provides a practical means by which to perform in-situ testing of post-frame buildings and decreases the complexity of post-frame building monitoring will contribute to the advancement of design and analysis techniques. In the calibration phase of the project, metal foil resistance strain gages were mounted onto wooden specimens with dimensions of 5 x 13 x 40 mm, 5 x 40 x 100 mm, and 2 x 20 x 50 mm, and acrylic specimens with dimensions of 3 x 25 x 75 mm. These specimens were then subjected to loading in an ATS universal testing machine in the Physical Properties Lab at the University of Manitoba. Stress-strain curves were developed based upon the observed stress and strain levels. These calibrated gages were then mounted on to a 38 x 89 mm specimen of S-P-F dimensional lumber which represented a typical light-framed building material. This assembly was then subjected to a similar loading procedure as the calibrated gage and stress-strain curves were generated once again. The slopes of the stress-strain curves developed from the two phases of the project were compared to determine if a consistent correlation existed. The three sizes of wood specimens did not demonstrate a consistent correlation. However, the acrylic specimen demonstrated consistent correlation amongst two groups of three with correlation coefficients within a forty percent range in one group and within a nine percent range in the other group. This suggests that further experimental refinements could produce the desired results. / October 2007

To be able to ensure quality, efficiency and safety in nuclear reactors, non-destructive evaluations (NDE) are performed. The moulding technique, which has been studied in this project, is an NDE method used to verify surface breaking cracks at various objects in reactor vessels. The idea of moulding is to receive a copy of the replicated surface for microscopic analysis. Within forensic science the moulding technique is used at crime scenes to collect evidence and tie suspects to crimes. Underwater moulding, however, is a newly developed technique and WesDyne TRC is a pioneer in offering services within moulding for underwater purposes. This project was initiated by WesDyne TRC to further their knowledge within the moulding technology. In the project, studies have been made at three important parameters effect on cast quality using three different polymer compounds. Problems during moulding, such as crack detection failures and bubbles in the casts, raise the question whether the underwater moulding technique can be trusted to detect cracks. Results from the experiments led to a greater insight into the problem with receiving high quality casts during underwater moulding. Only if a satisfactory cast is made, the moulding method can be trusted to detect defects down to the detection target in both dry and wet environment. To increase the surface quality of underwater casts a suggestion for mould design and a recommended moulding method was developed. In addition, one of the polymer compounds approved for use, turned out not to be suitable for underwater moulding.

The heterogeneous and anisotropic nature of wood material creates additional design challenges not present with the use of other structural materials such as steel and aluminum. The natural variation in the physical properties of wood members requires that the specified strengths and resistances used for design calculations be based on the quantities measured for the fifth percentile of all wood materials tested. The result is that design may be unnecessarily conservative and subsequently inefficient. The same properties that cause uncertainty surrounding the physical properties of biological materials also create difficulty in applying non-destructive evaluation techniques. Strain measurement is one particular technique that is extremely valuable for materials of known and consistent stress-strain relationships, but whose usefulness is diminished when applied to biological materials. To demonstrate the need for more accurate strain measurement in light-framed structures, the predictive calculations and structural modelling of a post-framed building was compared to its demonstrated performance. The analysis did not adequately reflect the actual performance of the building, and it was determined that additional monitoring of light-framed buildings through systems such as strain measurement was required to gain a better understanding of the performance characteristics in order to optimize evaluation techniques. This project aimed to develop a system that accurately measures strain in dimensional lumber of different types, which in turn will enable researchers to enhance monitoring the performance of light-frame structures and optimize design analysis and structural modelling techniques. The development of a methodology that provides a practical means by which to perform in-situ testing of post-frame buildings and decreases the complexity of post-frame building monitoring will contribute to the advancement of design and analysis techniques. In the calibration phase of the project, metal foil resistance strain gages were mounted onto wooden specimens with dimensions of 5 x 13 x 40 mm, 5 x 40 x 100 mm, and 2 x 20 x 50 mm, and acrylic specimens with dimensions of 3 x 25 x 75 mm. These specimens were then subjected to loading in an ATS universal testing machine in the Physical Properties Lab at the University of Manitoba. Stress-strain curves were developed based upon the observed stress and strain levels. These calibrated gages were then mounted on to a 38 x 89 mm specimen of S-P-F dimensional lumber which represented a typical light-framed building material. This assembly was then subjected to a similar loading procedure as the calibrated gage and stress-strain curves were generated once again. The slopes of the stress-strain curves developed from the two phases of the project were compared to determine if a consistent correlation existed. The three sizes of wood specimens did not demonstrate a consistent correlation. However, the acrylic specimen demonstrated consistent correlation amongst two groups of three with correlation coefficients within a forty percent range in one group and within a nine percent range in the other group. This suggests that further experimental refinements could produce the desired results.

The heterogeneous and anisotropic nature of wood material creates additional design challenges not present with the use of other structural materials such as steel and aluminum. The natural variation in the physical properties of wood members requires that the specified strengths and resistances used for design calculations be based on the quantities measured for the fifth percentile of all wood materials tested. The result is that design may be unnecessarily conservative and subsequently inefficient. The same properties that cause uncertainty surrounding the physical properties of biological materials also create difficulty in applying non-destructive evaluation techniques. Strain measurement is one particular technique that is extremely valuable for materials of known and consistent stress-strain relationships, but whose usefulness is diminished when applied to biological materials. To demonstrate the need for more accurate strain measurement in light-framed structures, the predictive calculations and structural modelling of a post-framed building was compared to its demonstrated performance. The analysis did not adequately reflect the actual performance of the building, and it was determined that additional monitoring of light-framed buildings through systems such as strain measurement was required to gain a better understanding of the performance characteristics in order to optimize evaluation techniques. This project aimed to develop a system that accurately measures strain in dimensional lumber of different types, which in turn will enable researchers to enhance monitoring the performance of light-frame structures and optimize design analysis and structural modelling techniques. The development of a methodology that provides a practical means by which to perform in-situ testing of post-frame buildings and decreases the complexity of post-frame building monitoring will contribute to the advancement of design and analysis techniques. In the calibration phase of the project, metal foil resistance strain gages were mounted onto wooden specimens with dimensions of 5 x 13 x 40 mm, 5 x 40 x 100 mm, and 2 x 20 x 50 mm, and acrylic specimens with dimensions of 3 x 25 x 75 mm. These specimens were then subjected to loading in an ATS universal testing machine in the Physical Properties Lab at the University of Manitoba. Stress-strain curves were developed based upon the observed stress and strain levels. These calibrated gages were then mounted on to a 38 x 89 mm specimen of S-P-F dimensional lumber which represented a typical light-framed building material. This assembly was then subjected to a similar loading procedure as the calibrated gage and stress-strain curves were generated once again. The slopes of the stress-strain curves developed from the two phases of the project were compared to determine if a consistent correlation existed. The three sizes of wood specimens did not demonstrate a consistent correlation. However, the acrylic specimen demonstrated consistent correlation amongst two groups of three with correlation coefficients within a forty percent range in one group and within a nine percent range in the other group. This suggests that further experimental refinements could produce the desired results.

The research work focuses on implementing intelligent measurement and diagnostic techniques for the non-destructive evaluation (NDE) of aircraft carbon composites. The outcome of this research work developed reliable and faster techniques to aid in the rapid assessment of defects in anisotropic carbon composites by applying ultrasonic and infrared thermography NDE methods. To fulfill the requirement of the intelligent non-destructive evaluation methods, this research is divided into four sub-researches: fuzzy logic based delamination detection, super-resolution image reconstruction for ultrasonic C-scan, ultrasonic 3D reconstruction, and polynomial fitting techniques for infrared thermography inspection. These researches focus on the improvement and optimization of current ultrasonic testing and infrared thermography inspection. They are independent but interrelated component, and they all serve the same goal which is to interpret data correctly and provide detailed information about the region of interests (ROI) for intelligent non-destructive measurement and evaluation. Details of these researches are presented in Chapter 2, 3, 4, and 5 respectively. For the ultrasonic testing, a fuzzy inference classifier will be used to generate the rule base and knowledge base for different kinds of defects in composites. It will automatically manage large amounts of signal data sets and extract the important information. Data features and NDE expert knowledge are seamlessly combined to provide the best possible diagnosis of the potential defects and problems. As a result, the outcome of this research work will help ensure the integrity and reliability of carbon composites. The C-scan image resolution of ultrasonic testing system was improved by applying super-resolution algorithms to overcome the inherent resolution limitations of the existing ultrasonic system. It greatly improves the image quality and allows for more detailed inspection of the ROI with high resolution, making defect evaluation easier and more accurate. The ultrasonic 3D reconstruction technique will be able to provide NDE inspectors with more detailed information on defect depth, volume, and 3D structure, as well as help them make quick, accurate, and reliable decisions. For the IR inspection, the thermography methods based on the thermal contrast are strongly affected by non-uniform heating which due to the heat source alignment and specimen thickness variation. The proposed polynomial curve fitting and surface fitting techniques were applied to eliminate the non-uniform heating effect by subtracting the estimated non-uniform heating pattern from the corrupted IR images. Mainly, aircraft composite material: carbon fiber reinforced polymer (CFRP) panels will be considered for this research work. Based on the preliminary study, delamination defects due to impact damage and foreign object inclusions artificially embedded in CFRP panels were successfully detected by immersion ultrasonic testing (UT) and IRT inspection. Therefore, the next step will be in improving the detection algorithm and developing an intelligent quality inspection technique for NDE testing. Powered with multiple image processing techniques and mathematical algorithms, the research result will provide high resolution images and detailed information about defect areas. In addition, it will also capable of identifying the type, shape, size, and the distribution of defect.

Deterioration of bridge decks is an ongoing problem faced by transportation agencies across the country. In past decades, Non-Destructive Evaluation (NDE) techniques, capable of detecting various deteriorations types, e.g., cracking, delamination and reinforcing steel corrosion, have emerged. These techniques generate large amounts of data representing different underlying physics, (decibels for ground penetrating radar and volts for half-cell potential), making data interpretation and comparison difficult for bridge owners and practitioners. The deterioration threshold, or the transition between healthy and deteriorated areas, is essential in understanding NDE data. However, this threshold is determined empirically in former research and engineering practice. In the present research, a probability-based method is proposed to identify deterioration thresholds for specified confidence levels. NDE data measuring different underlying physics are transformed into a binary format by threshold values to compare and combine multiple NDE techniques for bridge deck assessment. The finite element method is also implemented to correlate bridge deck surface stresses with deteriorations measured by NDE techniques, and to study the causes on concrete bridge deck degradation. The general methodology developed in this study will be demonstrated on three bridges, i.e., Virginia, New Jersey and New York Pilot Bridges, which were studied under Federal Highway Administration (FHWA)'s Long-Term Bridge Performance (LTBP) Program. / Ph. D.

La méthode des courants de Foucault (CF) est très largement utilisée en milieu industriel pour l’évaluation non destructive (END) de pièces ou de structures électriquement conductrices, parce qu’elle est sensible, robuste peu couteuse et non polluante. Toutefois, dans le cas général, l’évaluation quantitative des paramètres caractéristiques d’une structure est un problème difficile, d’une part parce que les données CF disponibles sont généralement incomplètes, d’autre part parce qu’il est nécessaire de faire appel à des modèle numériques élaborés pour rendre compte des interactions sonde / structure, et enfin, parce que le problème inverse consistant à estimer les paramètres de la structure à partir de la connaissance des interactions physiques et des données CF disponibles est « mal posé ».Dans cette thèse, nous nous intéressons au problème particulier de l’évaluation de jeux entre pièces dans un assemblage métallique. Pour contourner les difficultés liées à la mise en oeuvre de l’évaluation quantitative par CF, nous avons choisi de placer nos travaux dans un cadre multifréquence afin d’enrichir les données d’observation fournies par la sonde, et d’adopter une démarche consistant à élaborer des modèles comportementaux issus de l’analyse statistique des interactions sonde / structure. Partant de l’analyse de l’effet d’un jeu apparaissant entre les plaques d’un assemblage de plaques d’aluminium sur la variation de l’impédance normalisée de la sonde, menée expérimentalement et par simulations à l’aide de codes éléments finis, nous avons tout d’abord bâti un modèle comportemental approché d’interactions reposant sur la linéarité de la relation observée dans une bande de fréquences d’examen liée aux dimensions de la structure étudiée. Dans ce cadre, nous avons proposé, implanté et discuté des performances de plusieurs algorithmes de résolution du problème inverse, permettant d’estimer le jeu entre pièces et l’épaisseur de la pièce « cachée » de l’assemblage. Ensuite, nous avons étendu l’approche proposée à l’aide de réseaux de neurones artificiels, puis appliqué la méthode comportementale proposée à un second problème : celui de la caractérisation de fissures de fatigue réelles dans des pièces massives. Les résultats obtenus inclinent à considérer que la démarche proposée est généralisable à d’autres configurations en END par CF. / The eddy current (EC) method is widely used in the industry for the nondestructive evaluation (NDE) of electrically conductive structures, because it is sensitive, robust, cost effective and non polluting. However, in the general case, the quantitative evaluation of the parameters of a structure is a difficult problem, because the available EC data are generally incomplete, because it is necessary to have numerical models developed to account for the probe / structure interactions, and finally, because the inverse problem which consists in estimating the parameters of an inspected structure from the knowledge of those physical interactions and the available EC data is "ill-posed".In this work, we focus on the evaluation of air gaps between the parts of a metallic assembly. In order to cope with the difficulties associated with the implementation of an EC quantitative evaluation, we choose to use a multi-frequency approach so as to increase the data provided by the EC probe, and to build behavioral models from the statistical analysis of the probe / structure interactions. Experimental data and simulated data based on finite elements modeling are analyzed. A first approach behavioral model of the probe/structure interactions is deduced from these analyses, which is based on the linear relationship observed (in a particular excitation an frequency band depending on the dimensions of the studied structure) between the variations of the sensor normalized impedance and the air gap to evaluate. Moreover, we propose, implement and discuss the performances of several algorithms designed to solve the inverse problem dealing with the estimation of both the multilayered structure air gap and "hidden" plate thicknesses. Then, we extend the proposed multi-frequency behavioral approach thanks to artificial neural networks, and we apply the proposed behavioral method to a second problem : that of the characterization of real fatigue cracks in metallic massive parts. The obtained results let the generalization of the proposed approach to other EC NDE configurations to be envisaged.

Gradient Enhanced Piezoelectric Actuators (GEPAC) are thin piezoelectric plates embedded between two composites layers having different thermal properties. Compared to standard unimorph bending actuators, GEPACs offer superior performances for operations at low frequencies. Potential applications are in the area of multifunctional aircraft skins. In practice, delaminations or debonding within the actuator itself can occur, and it is highly desirable to develop an ultrasonic nondestructive method to monitor the integrity of the actuator in real time. For this study, the composite material is unidirectional Kevlar-epoxy, with fibers oriented at 90 and 0 for the upper and lower layers to achieve different coefficient of thermal expansion. A thin PZT plate is inserted between the two layers, and extended copper foil is used for electrodes on the PZT. The first objective of the research is to demonstrate that, by using segmented electrodes, one can simultaneously launch an ultrasonic pulse (1 MHz) for NDE testing while the actuator is undergoing low frequency actuation (less than 100 Hz). The second objective is to show that the ultrasonic signal can be used to detect damage induced during fatigue testing of the actuator. The third objective is to use the technique to monitor the integrity of a composite plate containing several embedded GEPACs.