<p> In this thesis we develop and evaluate automated "expert systems" for two applications: (i) gas/oil pipeline inspection using magnetic flux leakage information, (ii) treatment efficacy prediction and medical diagnosis using electroencephalograph (EEG) and clinical information. Both applications share the same methodology and procedure as they employ machine learning methods which learn their decision models using the training data (or past examples in real life/environment).</p> <p> The magnetic flux leakage (MFL) technique is commonly used for nondestructive testing (NDT) of oil and gas pipelines which are mostly buried underground. This testing involves the detection of metal defects and anomalies in the pipe wall, and the evaluation of the severity of these defects. The difficulty with the MFL method is the extent and complexity of the analysis of the MFL images. In this thesis we show how modern machine learning techniques can be used to considerable advantage in this respect.</p> <p> The problem of identifying in advance the most effective treatment agent for various psychiatric conditions remains an elusive goal. To address this challenge, an automated medical expert system is designed and then evaluated. The system is capable of predicting the treatment response for each individual patient at the outset of a therapy (i.e., using pre-treatment information) thus improving therapeutic efficiency and reducing personal and economic costs. Our experiments are focused on treatment planning and diagnosis of mood disorders and psychiatric illnesses. Through different experiments, we have shown that it is possible to predict treatment efficacy of a 'selective serotonin reuptake inhibitor' (SSRI) antidepressant and 'repetitive transcranial magnetic stimulation' (rTMS) therapies for patients with treatment-resistant major depressive disorder (MDD) or major depression. The predictions are based on pre-treatment quantitative EEG measurements. Also, prediction of post-treatment schizophrenia symptomatic scores, using pre-treatment EEG data, showed significant performance in patients treated with the drug clozapine. Clozapine is an antipsychotic medication of superior effectiveness in treating Schizophrenia but has several potentially severe side effects.</p> <p> Medical diagnosis is the second problem we consider in the neuroscience aspects of this thesis. In this research, an automated digital medical diagnosis methodology is developed to estimate/detect the type of a disease or illness that a patient is suffering. This intelligent diagnostic system can assist the physician/clinician by offering a second opinion on diagnosis. Several complex psychiatric illnesses may have many common symptoms and accurate diagnosis can, at times, be very difficult. Efficient diagnosis helps by avoiding prescription of wrong therapy /treatment to a patient. In our limited experiments, EEG data is used to make a diagnosis for distinguishing between various psychiatric illnesses including MDD, schizophrenia, and the depressed phase of bipolar affective disorder (BAD).</p> <p> In all problems considered in this thesis, specifically the neuroscience problem, a large number of candidate features are extracted from measurement data but most candidate features are found to be irrelevant and have little or no discriminative power. Finding a few most discriminating features that guarantee numerical efficiency and obtain a smooth and generalizable decision function, is a major challenge in this research. In this thesis, feature selection methods based on mutual information or Kullback-Leibler (KL) distance is employed to find the most statistically relevant features. For the multi-class diagnosis problem, to improve performance, a feature selection procedure denoted as feature combination feature selection is used which first finds discriminating features in all binary classification combinations, and then combines them into a larger feature subset to make a final multi-class decision. The two-dimensional (2D) representation of the feature data is also found to be useful for clustering analysis. The overall method was evaluated using a nested cross-validation procedure for which over 80% average prediction performance is obtained in all experiments. The results indicate that machine learning methods hold considerable promise in solving the challenging problems encountered in the two applications of concern.</p> / Thesis / Doctor of Philosophy (PhD)

<p> Software Pipelining is a method of instruction scheduling where loops are scheduled more efficiently by executing operations from more than one iteration of the loop in parallel. Finding an optimal software pipelined schedule is NP-complete, but many heuristic algorithms exist. </p> In iteration i, a software pipelined loop will execute, in parallel, "stage" 1 of iteration i, stage 2 of iteration i- 1 and so on until stage k of iteration i-k+l. </p> <p> We present a new approach to software pipelining based on using a hemistic algorithm to explicitly assign each operation to its stage before the actual scheduling. </p> <p> This explicit assignment allows us to implement control flow mechanisms that are hard to implement with traditional methods of software pipelining, which do not give us direct control over what stages instructions are assigned to. </p> / Thesis / Master of Science (MSc)

No description available.

The aim of this research is to identify problematic microstructural features as hydrogen traps in linepipe steel that serve to increase the hydrogen embrittlement susceptibility. A comparison is made between the hydrogen trapping capacity and associated hydrogen embrittlement susceptibility of Ca-treated X60 grade steel skelp and X70 grade steel skelp: the latter typically being more susceptible to hydrogen-induced cracking in sour environments. Through-thickness variations in the steel skelp microstructure were characterized across multi-length scales using light optical microscopy (LOM) and scanning electron microscopy (SEM) equipped with X-ray energy dispersive spectroscopy (EDS). Key features under study include the composition, shape, and distribution of non-metallic inclusions, as well as differences in features present between the quarterline (¼ and ¾ depths) and centerline (½ depth) microstructures. The type, count, and average size of inclusions present in both steel skelp grades were analyzed using an automated SEM-EDS technique called ASPEX®. Major types of inclusions detected in both grades of steel skelp include those containing Ca, Al, Mn, Mg and Ti as major elements. Overall, the area fraction of inclusions detected in the X70 steel was larger than those detected in the X60 with the exception of Ti-containing inclusions, which had a larger area fraction within the X60 steel. Comparing the number of detected inclusions shows that there was overall slightly less Ca-containing inclusions and significantly less Ti-containing inclusions detected in the X70 steel but there was generally more Al-containing, Mg-containing, and Mn-containing inclusions than those detected in the X60 steel. Thermal desorption spectroscopy (TDS) measurements were made on samples prepared from the ¼, ½, and ¾ depths of X60 and X70 steel skelps after galvanostatic cathodic charging in an As2O3-containing solution using an applied current density of −10 mA/cm2. Hydrogen release was measured using a HYDROSEEL® probe while the sample was heated from 20°C to 650°C to detect temperature values at which hydrogen gas release peaks occurred, and thus provide information on types of reversible and/or irreversible traps present. The TDS results suggests that non-metallic inclusions indeed serve as irreversible traps along with grain boundaries and dislocations, which serve as reversible traps. Hydrogen permeation measurements were also made on samples prepared from the ¼, ½, and ¾ depths after galvanostatic cathodic charging in an As2O3-containing solution using an applied current density of −10 mA/cm2. Hydrogen gas release was measured using a HYDROSEEL® probe while the sample remained at room temperature (~20°C), providing information regarding the potency of reversible hydrogen traps when subjected to a flux of hydrogen. Only reversible traps can be detected at room temperatures due to their low binding energies. Higher temperatures are required to overcome the larger binding energies associated with irreversible traps. The hydrogen permeation results indicate no significant effect of through-thickness variations in the X60 steel, but the centreline depth of the X70 steel skelp trapped a larger quantity of hydrogen than either of the two quarterline depths, indicating the presence of a distinct problematic trap. The X70 steel skelp was also observed to trap more hydrogen than the X60 steel skelp. The observed hydrogen trapping capacity was linked to the hydrogen embrittlement susceptibility by comparing the uniaxial tensile behaviour of centreline samples with and without hydrogen charging applied as a pre-treatment step. Hydrogen charging was achieved by galvanostatic cathodic polarization at an applied current density of −10 mA/cm2 for 24 h in an NH4SCN-containing solution while simultaneously loading the samples to 85% of the yield strength using a proof ring tensile test cell. An increase in hydrogen embrittlement as a result of pre-charging was confirmed through tensile plots by comparing the area of reduction and failure strain of charged samples to uncharged samples. A decrease in both values was observed in the charged samples indicating a loss in ductility as a result of hydrogen charging. Fracture surfaces were imaged using SEM and inclusions of interest were analyzed for elemental composition using EDS. Inclusions observed along the fracture surfaces include oxysulfides of Ca and Al, oxides of Mg, Al-Ca-Si oxides, and Al2O3-containing inclusions which are likely to be heterogeneous Al-Ca-O inclusions. / Thesis / Master of Applied Science (MASc)

With new government regulations and emerging knowledge of the risk to the environment posed by the failure of stormwater pipelines, stormwater infrastructure asset management is becoming increasingly important in the U.S. An essential aspect of asset management practice is the accurate performance assessment of one's assets. This paper presents a weighted factor framework to determine the performance of stormwater pipes. This paper prepares a list of 50 parameters affecting the performance of stormwater pipelines; the list is based on a review of the literature, existing asset management plans, and feedback from utilities. This list is broken down into essential and preferential parameters. Indeed, not all utilities necessarily possess sufficient resources to collect such a large set of parameters. This study also develops a three-level hierarchical structure of the degradation of stormwater pipeline infrastructure. The structure consists of five failure modules and the essential parameters only. On the basis of the survey results gathered from 10 utilities across the EPA regions, the study combines the essential parameters into a performance index. The index is a scale of 1 to 5, similar to the National Association of Sewer Service Companies' Pipeline Assessment and Certification Program grading system. Grade 1 implies excellent condition and Grade 5 implies collapse is imminent. / Master of Science

A single compressor station site along a gathering line network was monitored for fugitive methane emissions to quantify long-term emissions in Appalachia Virginia. Continuous monitoring was conducted from January 2021 to April 2021. The compressor station undergoing monitoring operated two CAT3516 Tale and one CAT3516 B engines operating at 80% of max output flow. Data presented on methane emissions during this period was gathered with an eddy covariance monitoring station. This station was equipped with an LI-7700 methane analyzer, LI-7500A - CO_2/H_2 O analyzer as well as a sonic anemometer these sensors could be observed remotely through cellular connection. The data is represented in flux output ((µmol)/(s m^2 )) as well as kg CO_2 equivalence of methane outlined by the EPA greenhouse gas inventory. The average daily emissions for this compressor station are estimated to be 136 kg CO_2 equivalent emissions. This study shows that the site during the observational period the compressor station emitted on average are estimated to be 5.43 kg of CH_4 per day. / Master of Science / There has been an increased interest in quantifying and recording methane (CH_4) emissions among all sectors. A main focus of interest among methane is to understand fugitive gasses and emissions resulting from the natural gas sector. Leaks along pipelines are most likely occurring at connection points between components. This study aimed to continuously monitor a pipeline compressor station in Appalachia Virginia. Compressor stations are just one component of the pipeline network as well as the natural gas production and delivery chain attributed with CH_4 emissions. To monitor methane emissions at the site a stationary eddy covariance monitoring station was installed that was equipped with an open path methane analyzer, open path CO_2 and H_2 O analyzer, and a sonic anemometer. The data gathered was used to calculate the flux of methane which is the amount of methane being generated or absorbed by the area of interest. The goal of this study was to continuously monitor methane emissions of a natural gas compressor station. Data presented in this study was collected from January 2021 to April 2021. Data was presented in the flux output ((µmol)/(s m^2 )) as well as kg CO_2 equivalence of methane outlined by the EPA greenhouse gas inventory.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.

Le but de notre étude était de mettre en place une méthode qui puisse nous permettre d’identifier, en aveugle, des perturbateurs de voies biologiques, et qui soit d’une part généralisable pour toute espèce de poissons et d’autre part applicable quel que soit le cours d’eau considéré. Nous nous sommes intéressés aux gènes différentiellement exprimés dans le foie, en utilisant la technologie du séquençage Illumina de banques ADNc. Nous avons étudié trois espèces de cyprinidae (C. nasus, P. toxostoma, S. cephalus) dans le bassin de la Durance, servant à l’alimentation en eau de plusieurs millions de personnes. Nous avons mis en place une suite logicielle pour inférer un transcriptome pour chacune des trois espèces étudiées, et effectué un travail en bioinformatique pour l’identification des spécimens hybrides. Cette méthode nous permet d’assigner les 596 millions de séquences générées (293 spécimens) à l’une des trois espèces et à 16606 gènes identifiés. Les résultats biologiques montrent des variations de l’expression de gènes touchant des voies biologiques associées à des réponses aux xénobiotiques le long de l’axe amont-aval de la rivière. Ils montrent aussi que les spécimens échantillonnés dans le canal EDF présentent des réponses atténuées aux xénobiotiques par rapport aux individus en rivière. Ce résultat peut s’expliquer par l’effet de dilution des polluants dans une masse d’eau plus importante. Cette étude met en évidence les capacités adaptatives des populations de poissons à court terme, via des modifications de l’expression des gènes à un ensemble de perturbateurs environnementaux. Ce travail permet d’envisager la mise en place d’un outil de gestion incontournable. / The aim of our study was to establish a method that allows the identification (in blind) of biologicalpathway disrupters, for all species of fish and applicable regardless of the watercourse considered. Wefocused on differentially expressed genes (and the biological pathways in which they act) in the liver,using the Illumina sequencing technology of cDNA libraries. We studied three species of cyprinidae (C.nasus, P. toxostoma, S. cephalus) in the Durance basin that constitutes water resource for several millionpeople. We have implemented a pipeline to infer the transcriptome for each of the three species studied,and developed a bioinformatics framework for the identification of hybrid specimens. This methodallows us to assign the 596 million sequences generated (representing 293 specimens) to one of the threespecies and to 16,606 identified genes. The biological results display variations in the expression of genesaffecting biological pathways associated with xenobiotic responses (estrogens, Hap, heavy metals) alongthe upstream-downstream axis of the river. They also yield that the specimens sampled in the EDFchannel displayed an attenuated responses to xenobiotics, in comparison to individuals that inhabitethe river, possibly a benefit of the dilution effect of pollutants in a larger body of water. This studyhighlights short-term adaptive capacities (acclimation) of fish populations to a set of environmentaldisrupters via changes in gene expression levels. It will open a way to an essential tool for managementpolicies.

A morfologia matemática é o estudo da forma utilizando as ferramentas da teoria de conjuntos e representa uma área extremamente importante em análise de imagens. Suas operações básicas são a dilatação e a erosão, e através destas é possível realizar outras operações mais complexas. A morfologia matemática fornece ferramentas poderosas para a realização de análise de imagens em baixo nível e tem encontrado aplicações em diversas áreas, tais como: visão robótica, inspeção visual, medicina, análise de textura, entre outras. Muitas destas aplicações requerem processamento em tempo real, e para sua execução de forma eficiente freqüentemente é utilizado hardware dedicado. A análise de imagens em baixo nível geralmente envolve computações repetidas sobre estruturas grandes de dados. Assim, o paralelismo parece ser um atributo necessário de um sistema de hardware capaz de executar eficientemente estas tarefas. As ferramentas da morfologia matemática são bem adequadas à implementação em arquiteturas pipeline. A necessidade de sistemas capazes de realizar o processamento de imagens digitais em tempo real, com o menor custo e tempo de desenvolvimento, tem sido suprida pela tecnologia de dispositivos de lógica programável complexa. Assim, neste trabalho foi projetada e implementada uma arquitetura pipeline dedicada para dilatação e erosão de imagens binárias em tempo real utilizando dispositivos lógicos programáveis de alta capacidade. Esta arquitetura é capaz de processar imagens binárias de 512 x 512 pixels. Os estágios desta arquitetura são flexíveis, permitindo a reprogramação da forma e do tamanho dos elementos estruturantes utilizados nas operações morfológicas. A arquitetura desenvolvida apresentou um desempenho satisfatório, demonstrando ser uma alternativa viável e eficiente. / Mathematical morphology is a very important image analysis area that uses set theory tools to study shapes. The basic operations in mathematical morphology are dilation and erosion, these can be used for more complex operations. Mathematical morphology has powerful tools for low level image processing and has been used in a wide range of applications such as robotic vision, visual inspection, medicine and texture analysis. Low level image processing requires repetitive processing over large data structures, dedicated parallel computing hardware is often used. Complex field programmable logic devices (CPLDs) have increasingly been used for the fast development of real time image processing systems. In this work we present a pipeline architecture for real time erosion and dilation operations, the architecture was developed using high density programmable logic devices. The developed architecture can process 512 x 512 pixels binary images, and has flexible stages that can be reprogrammed according to the shape and size of the structuring elements used in the morphological operations. Tests performed using the architecture demonstrated its good performance and that it is a good and efficient alternative for dedicated morphological image processing operations.

Anotação proteogenômica é uma abordagem que une a análise proteômica com a anotação genômica. O intuito de tal abordagem é prover uma anotação mais detalhada ao gene. Intuito esse, que nem sempre é possível quando se trata apenas de genes, uma vez que produtos gênicos, com funções importantes preditas, somente passam a ter papel na fisiologia do organismo quando expressos e traduzidos. Com todo o avanço atual de estudos na área proteogenômica, a geração de dados tem crescido de modo exponencial e, com esse crescimento, nota-se a necessidade cada vez maior da criação de sistemas capazes de processar, armazenar e gerenciar essas novas informações produzidas. Assim, é descrito nesse trabalho o desenvolvimento do ProGen AP , sendo constituído de uma interface web construída em HTML/PHP5, um banco de dados cujo SGBD é o mySQL e de módulos de processamento de dados proteômicos, neste caso o LabKey (com o core Xtandem!) e o QuickMod. Todos os módulos são open source e comunicam entre si através de scripts PERL. Nesse sistema, o pesquisador fornece dados de experimentos proteômicos e o sistema, então, os processa e retorna ao usuário informações sobre o gene expresso, a localização dos peptídeos dentro do gene aos quais pertencem e, ainda, informações quantitativas sobre o peptídeo e a proteína identificados. Além disso, o uso de um processamento esquematizado reduz a possibilidade de erro de entrada/saída de dados nos módulos intermediários do processamento. Aqui, o ProGen AP foi aplicado no estudo proteômico do Mycobacterium tuberculosis (MTb). Na literatura, o genoma do MTb cepa H37Rv contém apenas 4062 open reading frames (ORFs) preditos e o complemento funcional desse genoma, o proteoma, ainda não está totalmente elucidado. A análise do proteoma do MTb, com o uso do ProGen AP, resultou em uma lista total de 154.982 identificações de peptídeos, representando um total de 147.334 peptídeos únicos. Até o momento, foram identificadas 2.369 proteínas, cobrindo aproximadamente 58% de todo o genoma do MTB. É importante ressaltar que, dentre todas as proteínas identificadas até o momento, a maioria delas está anotada como proteinas hipotéticas em seu genoma, e, por consequência, os resultados obtidos nesse projeto confirmam e validam a existência de tais produtos gênicos. Além disso, 567 peptídeos foram identificados como N-terminal e 1229 como C-terminal, o que indica a correta predição do início e do término da tradução de tais genes. Todos esses resultados positivos confirmam que a abordagem utilizada no ProGen AP é eficiente e pode ser usada em vários outros organismos de interesse do pesquisador. / Proteogenomic annotation is an approach that combines proteomic analysis and genomic annotation. The aim of this approach is to provide a more detailed annotation, which is not possible in most of the times when dealing mostly with genes, once that genomic products, with important predicted functions are only important in the organism physiology when they are expressed and translated. There have been occurring several advances in proteogenomic studies and the generation of new data sets has been growing in an exponential wave. With all this growth, the creation of systems able to storing, processing and analyzing all the new knowledge produced is eminent. This study presents the deployment of ProGen AP, a system built with a HTML/PHP5 web interface, a mySQL data management system to store the data and two processing modules (LabKey, with core X!Tandem and QuickMod). In this system, the researcher provides a data set from a proteomic experiment and then the system processes it and returns to the researcher information about the expressed gene, the peptides localization inside the gene that they belong and, also, quantitative information about the peptide and the protein that were identified. Also, the use of an automated pipeline reduces the possibility of making mistakes in input/output of the data when using the intermediate modules. Here, the ProGen AP were applied to perform a proteogenomic annotation of Mycobacterium tuberculosis (MTb). In literature, the MTb genome, strain H37RV, have only 4062 predicted open reading frames (ORFs) and the functional complement of this genome is not completely known. The MTb analysis using ProGen AP, resulted in a list of 154.982 peptides identification, representing a total of 147.334 single peptides. Until now, were identified 2.369 proteins, covering nearly of 58% of the whole MTb genome. Is very important to highlight that, among all the identified proteins until now, most of them are annotated as hypothetical proteins in the MTb genome, so can be affirmed that the results of this project can confirm and validate the existence of all these genomic products. Beside this, 567 peptides were identified as been an N-terminal peptide and 1229 were identified as been a C-terminal, this fact indicates that the prediction of the beginning and the end of translation of those genes are right. All these positive results corroborate that the approach utilized in the ProGen AP is efficient and can be used in studies of other organisms.