Currently, water utilities are facing monetary crises to maintain and expand services to meet the current as well as the future demands. Standard practice in pipeline infrastructure asset management is to collect data and predict the condition of pipelines using models and tools. Water utilities want to be proactive in fixing or replacing the pipes as fixing-when-it-fails ideology leads to increased cost and can affect environmental quality and societal health. There is a number of modeling techniques available for assessing the condition of the pipelines, but there is a massive shortage of methods to check the reliability of the results obtained using different modeling techniques. It is mainly because of the limited data one utility collects and absence of piloting of these models at various water utilities. In general, water utilities feel confident about their in-house condition prediction and failure models but are willing to utilize a reliable methodology which can overcome the issues related to the validation of the results. This paper presents the methodology that can enhance the reliability of model results for water pipeline performance analysis which can be used to parallel the output of the real system with confidence. The proposed methodology was checked using the dataset of two large water utilities and was found that it can potentially help water utilities gain confidence in their analyses results by statistically signifying the results. / Master of Science

The diploma thesis deals with the sphere of solder joints reliability. The narrower focus is use of lead-free solder alloy SAC305 in the production process and parameters of its reliability. The text describes the main factors, which have the influence on the reliability of solder joints under conditions of thermal cycling. These factors also relate with choice of substrates and technological processes of preparation, which are characterized and described. Another part is devoted to estimating the reliability of soldered connections and are listed the fatigue model to estimate reliability. These fatigue models are categorized based on different physical mechanisms that operate in the soldered joints during operation. Based on the comparison of different models is selected the most appropriate model and in conjunction with simulation in ANSYS is estimated reliability. For this purpose is selected soldered connection of the FR-4 substrate and ceramic substrate via SMD component. They are manufactured test kits and subjected to conditions of temperature cycling. Results obtained from experimental measurements are compared with results obtained by simulation and calculation.

Artificial intelligence (AI) and machine learning (ML) present some application opportunities and challenges that can be framed as learning problems. The performance of machine learning models depends on algorithms and the data. Moreover, learning algorithms create a model of reality through learning and testing with data processes, and their performance shows an agreement degree of their assumed model with reality. ML algorithms have been successfully used in numerous classification problems. With the developing popularity of using ML models for many purposes in different domains, the validation of such predictive models is currently required more formally. Traditionally, there are many studies related to model evaluation, robustness, reliability, and the quality of the data and the data-driven models. However, those studies do not consider the concept of the applicability domain (AD) yet. The issue is that the AD is not often well defined, or it is not defined at all in many fields. This work investigates the robustness of ML classification models from the applicability domain perspective. A standard definition of applicability domain regards the spaces in which the model provides results with specific reliability. The main aim of this study is to investigate the connection between the applicability domain approach and the classification model performance. We are examining the usefulness of assessing the AD for the classification model, i.e. reliability, reuse, robustness of classifiers. The work is implemented using three approaches, and these approaches are conducted in three various attempts: firstly, assessing the applicability domain for the classification model; secondly, investigating the robustness of the classification model based on the applicability domain approach; thirdly, selecting an optimal model using Pareto optimality. The experiments in this work are illustrated by considering different machine learning algorithms for binary and multi-class classifications for healthcare datasets from public benchmark data repositories. In the first approach, the decision trees algorithm (DT) is used for the classification of data in the classification stage. The feature selection method is applied to choose features for classification. The obtained classifiers are used in the third approach for selection of models using Pareto optimality. The second approach is implemented using three steps; namely, building classification model; generating synthetic data; and evaluating the obtained results. The results obtained from the study provide an understanding of how the proposed approach can help to define the model’s robustness and the applicability domain, for providing reliable outputs. These approaches open opportunities for classification data and model management. The proposed algorithms are implemented through a set of experiments on classification accuracy of instances, which fall in the domain of the model. For the first approach, by considering all the features, the highest accuracy obtained is 0.98, with thresholds average of 0.34 for Breast cancer dataset. After applying recursive feature elimination (RFE) method, the accuracy is 0.96% with 0.27 thresholds average. For the robustness of the classification model based on the applicability domain approach, the minimum accuracy is 0.62% for Indian Liver Patient data at r=0.10, and the maximum accuracy is 0.99% for Thyroid dataset at r=0.10. For the selection of an optimal model using Pareto optimality, the optimally selected classifier gives the accuracy of 0.94% with 0.35 thresholds average. This research investigates critical aspects of the applicability domain as related to the robustness of classification ML algorithms. However, the performance of machine learning techniques depends on the degree of reliable predictions of the model. In the literature, the robustness of the ML model can be defined as the ability of the model to provide the testing error close to the training error. Moreover, the properties can describe the stability of the model performance when being tested on the new datasets. Concluding, this thesis introduced the concept of applicability domain for classifiers and tested the use of this concept with some case studies on health-related public benchmark datasets. / Ministry of Higher Education in Libya

The Item response Theory and its Application in the Sociology Abstract The focus of this master's thesis is an item response theory, a modern methodological approach, which models a relationship between a latent variable and measured responses to the particular item. A theoretical introduction, which explains fundamentals of the theory, is followed by a comparison with the classical test theory and suggestions of its possible applications in a field of sociology. The thesis is completed with an illustrative example using real data to evaluate an instrument for measuring subjective quality of working life in the framework of item response theory. Keywords: Item response theory, scale, reliability, model, subjective quality of working life

Plusieurs organismes ont besoin d’information à propos de la valeur des alternatives afin de prendre la meilleure décision dans plusieurs contextes importants tels que l’approvisionnement, le choix d’un habitat ou d’un partenaire de reproduction. L’utilisation d’information sociale qui provient des congénères est généralement considérée comme avantageuse, dans la mesure où elle est peu couteuse à acquérir. Cependant, elle peut s’avérer néfaste lorsque celle-ci est incorrecte, désuète ou émise par un émetteur malhonnête dans le but de duper. Ainsi, deux questions cruciales se posent lorsque qu’un organisme a accès à de l’information sociale : quand copier ? et qui copier ? Parmi plusieurs autres facteurs, les capacités cognitives des individus pourraient affecter leur probabilité de copier et d’être copié. Spécifiquement, un individu ayant de bonnes capacités d’apprentissage serait plus certain de son information personnelle et devrait de ce fait moins utiliser l’information sociale qu’un individu moins performant. Les démonstrateurs ayant de meilleures aptitudes (en termes par exemple de capacité à résoudre de nouveaux problèmes) devraient également être considérés comme plus fiables et, de ce fait, être copiés préférentiellement. Afin de tester ces deux hypothèses, nous avons effectué une expérience avec des diamants mandarins (Taenyopigia guttata), dans laquelle les sujets pouvaient copier le choix de 2 démonstrateurs qui leur procuraient tous les deux une information sociale contradictoire à leur information personnelle mais différaient dans leur capacité à résoudre un problème. Les sujets étaient par la suite testés dans une tâche d’apprentissage associatif. À l’encontre de nos hypothèses, les individus plus performants dans la tâche d’apprentissage n’ont pas utilisé l’information sociale moins souvent que ceux qui ont eu besoin d’un grand nombre d’essais pour atteindre le critère d’apprentissage. Par contre, nous avons trouvé que les sujets qui avaient une préférence plus marquée utilisaient moins l’information sociale que ceux qui avaient fait un choix plus aléatoire. De plus, les individus qui ont utilisé l’information sociale ont copié plus fréquemment le choix du démonstrateur qui avait été observée en train de réussir une tâche de résolution de problème plutôt que celui qui avait échoué. Ainsi notre étude démontre que les diamants mandarins évaluent la pertinence de l’information personnelle et social en considérant leur niveau d’incertitude et les performances cognitives des démonstrateurs dans une tâche de résolution de problème. / Animals frequently rely on others’ choice to decide with whom to mate or what to eat. Because social information use may lead to maladaptive decisions, individuals should not rely on social information unconditionally. In particular, reliance on social information should be more frequent when uncertainty about how to behave is high, and hence in individuals with poor learning abilities. When individuals who are copied provide contradictory social information, social learners should also preferentially copy the decisions of the most reliable model, and so, for instance, better-performing models in cognitive tasks. Here we conducted an experiment with zebra finches (Taeniopygia guttata) on which we measured (i) the preference for three colored feeders before and after they had received conflicting information from two models that respectively succeeded or failed to resolve a novel problem and (ii) the performance on an associative learning task. We found no evidence that individuals that were quicker to learn a new association relied less on social information compared to slower learners. Yet, the birds who exhibited a stronger initial preference used social information to a lesser extent compared to those who showed a less pronounced preference. Furthermore, we found that individuals copied more frequently the choice of the problem-solver over the non-solver model. Thus, our findings demonstrate that zebra finches weight personal and social information based on their level of uncertainty and the models’ performance in cognitive tasks.

In the thesis, a methodology alternative to existing methods of digital systems design with increased dependability implemented into FPGA is presented, new features which can be used in the implementation and testing of these systems are demonstrated. The research is based on the use of FPGA partial dynamic reconfiguration for the design of fault tolerant systems. In these applications, the partial dynamic reconfiguration can be used as a mechanism to correct the fault and recover the system after the fault occurrence. First, the general principles of diagnostics, testing and digital systems dependability are presented including a brief description of FPGA components and their architectures. Next, a survey of currently used methods and techniques used for the design and implementation of fault tolerant systems into FPGA is described, especially the methods used for fault detection and localization, their correction, together with the principles of evaluating fault tolerant systems design quality.  The most important part of the thesis is seen in the description of the design methodology, implementation and testing of fault tolerant systems implemented into FPGAs which uses SRAMs as the configuration memory. First, the methodology of developing and automated checker components design for digital systems and communication protocols is presented. Then, a reference architecture of a dependable system implemented into FPGA is demonstrated including several fault tolerant architectures based on the use of partial dynamic reconfiguration as the mechanism of fault correction and the recovery from it. The principles of controlling the reconfiguration process are described together with the description of the test platform which allows to test and verify the design of fault tolerant systems based on the methodology presented in the thesis. The experimental results and the contribution of the thesis are discussed in the conclusions.

In the thesis, a methodology alternative to existing methods of digital systems design with increased dependability implemented into FPGA is presented, new features which can be used in the implementation and testing of these systems are demonstrated. The research is based on the use of FPGA partial dynamic reconfiguration for the design of fault tolerant systems. In these applications, the partial dynamic reconfiguration can be used as a mechanism to correct the fault and recover the system after the fault occurrence. First, the general principles of diagnostics, testing and digital systems dependability are presented including a brief description of FPGA components and their architectures. Next, a survey of currently used methods and techniques used for the design and implementation of fault tolerant systems into FPGA is described, especially the methods used for fault detection and localization, their correction, together with the principles of evaluating fault tolerant systems design quality.  The most important part of the thesis is seen in the description of the design methodology, implementation and testing of fault tolerant systems implemented into FPGAs which uses SRAMs as the configuration memory. First, the methodology of developing and automated checker components design for digital systems and communication protocols is presented. Then, a reference architecture of a dependable system implemented into FPGA is demonstrated including several fault tolerant architectures based on the use of partial dynamic reconfiguration as the mechanism of fault correction and the recovery from it. The principles of controlling the reconfiguration process are described together with the description of the test platform which allows to test and verify the design of fault tolerant systems based on the methodology presented in the thesis. The experimental results and the contribution of the thesis are discussed in the conclusions.