5G is currently being implemented around the world. A way to save resources in 5G could be to have several sector carriers sharing one power source. This requires being able to predict the electrical power of the sector carriers to make sure they do not exceed the capability of the power source. This thesis investigates deep learning for predicting the electrical power of a sector carrier. The chosen deep learning models are Multilayer perceptron (MLP) and Long short-term memory network (LSTM) and they are trained and evaluated on seven generated datasets from a sector carrier in Lund. The study consists of two parts, comparing the predictive performance between MLP and LSTM as well as determining the importance of each input feature on the predictions. It is concluded from the results that the MLP outperforms the LSTM in all datasets and should be the preferred model. The most important input feature for the predictions is by far the number of Resource elements (REs) used per slot. The number of Physical resource blocks (PRBs), modulation, scaling, number of streams and codebook indices have varying importance. The results can be used for guidance when implementing a real time system for which model and which input features to include. More work is required in investigating how the models perform on aggregated data from several sector carriers and determining if the models could feasibly be implemented for real time predictions. / 5G håller på att implementeras runtom i världen. Ett sätt att spara resurser i 5G skulle kunna vara att ha flera sektorbärare dela på en strömkälla. För att implementera detta behöver man kunna förutsäga effekten av sektorbärarna för att försäkra sig om att de inte kommer överstiga den maximala effekten strömkällan tillåter. I denna avhandling undersöks djupinlärning för att förutsäga effekten av en sektorbärare. De använda djupinlärningsmodellerna är Multilayer perceptron (MLP) och Long short-term memory (LSTM) och de tränas och utvärderas på sju stycken olika dataset genererade från en sektorbärare i Lund. Studien är uppdelad i två delar, en del där MLP och LSTM jämförs med hänsyn till deras prediktiva förmåga och en del där varje datavariabel utvärderas till hur viktig den är för modellernas förutsägelser. Från resultaten kunde slutsatsen dras att MLP presterar bättre än LSTM på alla sju dataset och att MLP bör väljas före LSTM. Den viktigaste datavariabeln är antalet resurs element med en signifikant skillnad jämfört med övriga datavariabler. Antalet fysiska resursblock, modulationen, skalningsfaktorn, antalet strömmar och kodboksindex har varierande viktighet. Resultaten kan användas som stöd för implementering av realtid system för att bestämma lämplig model och data. Mer arbete krävs för att undersöka hur modellerna presterar på data som är aggregerat från flera sektorbärare samt undersöka huruvida det är genomförbart att implementera ett realtid system för förutsägelser utifrån beräkningskraft och kostandsanalys.

<p>Streamflow prediction in ungauged basins (PUB) is a process generating streamflow time series at ungauged reaches in a river network. PUB is essential for facilitating various engineering tasks such as managing stormwater, water resources, and water-related environmental impacts. Machine Learning (ML) has emerged as a powerful tool for PUB using its generalization process to capture the streamflow generation processes from hydrological datasets (observations). ML’s generalization process is impacted by two major components: data splitting process of observations and the architecture design. To unveil the potential limitations of ML’s generalization process, this dissertation explores its robustness and associated uncertainty. More precisely, this dissertation has three objectives: (1) analyzing the potential uncertainty caused by the data splitting process for ML modeling, (2) investigating the improvement of ML models’ performance by incorporating hydrological processes within their architectures, and (3) identifying the potential biases in ML’s generalization process regarding the trend and periodicity of streamflow simulations.</p><p>The first objective of this dissertation is to assess the sensitivity and uncertainty caused by the regular data splitting process for ML modeling. The regular data splitting process in ML was initially designed for homogeneous and stationary datasets, but it may not be suitable for hydrological datasets in the context of PUB studies. Hydrological datasets usually consist of data collected from diverse watersheds with distinct streamflow generation regimes influenced by varying meteorological forcing and watershed characteristics. To address the potential inconsistency in the data splitting process, multiple data splitting scenarios are generated using the Monte Carlo method. The scenario with random data splitting results accounts for frequent covariate shift and tends to add uncertainty and biases to ML’s generalization process. The findings in this objective suggest the importance of avoiding the covariate shift during the data splitting process when developing ML models for PUB to enhance the robustness and reliability of ML’s performance.</p><p>The second objective of this dissertation is to investigate the improvement of ML models’ performance brought by Physics-Guided Architecture (PGA), which incorporates ML with the rainfall abstraction process. PGA is a theory-guided machine learning framework integrating conceptual tutors (CTs) with ML models. In this study, CTs correspond to rainfall abstractions estimated by Green-Ampt (GA) and SCS-CN models. Integrating the GA model’s CTs, which involves information on dynamic soil properties, into PGA models leads to better performance than a regular ML model. On the contrary, PGA models integrating the SCS-CN model's CTs yield no significant improvement of ML model’s performance. The results of this objective demonstrate that the ML’s generalization process can be improved by incorporating CTs involving dynamic soil properties.</p><p>The third objective of this dissertation is to explore the limitations of ML’s generalization process in capturing trend and periodicity for streamflow simulations. Trend and periodicity are essential components of streamflow time series, representing the long-term correlations and periodic patterns, respectively. When the ML models generate streamflow simulations, they tend to have relatively strong long-term periodic components, such as yearly and multiyear periodic patterns. In addition, compared to the observed streamflow data, the ML models display relatively weak short-term periodic components, such as daily and weekly periodic patterns. As a result, the ML’s generalization process may struggle to capture the short-term periodic patterns in the streamflow simulations. The biases in ML’s generalization process emphasize the demands for external knowledge to improve the representation of the short-term periodic components in simulating streamflow.</p>

No description available.

To combat anthropogenic climate change and comply with the Paris Agreement, the EU has previously introduced its Emissions Trading System (ETS) and has now also proposed a Carbon Border Adjustment Mechanism (CBAM). While these tools may reduce emissions within the EU, and in some cases, even in other countries, they can also affect European industries in unpredictable and sometimes negative ways. One industry that is currently included in the ETS is the steel industry. While the steel industry has not been signifi cantly affected yet, as the prices for ETS allowances remain relatively low, this is likely to change as the price for allowances rises. By identifying the factors which lead to cost differences between steel manufacturers in European countries through interviews and literature and comparing the costs of two manufacturers with similar production methods, this study aimed to quantify the cost diff erences that arose due to the ETS by constructing a model capable of estimating the cost differences. Furthermore, plausible changes to the ETS were identified and developed into scenarios where the cost differences that occurred between the two companies due to the new circumstances were analyzed. Energy was found to be the biggest cause of differences in cost, followed by transport costs if one company had a location advantage due to being closer to the customer. Historically, the cost of purchasing ETS allowances was nearly negligible. However, when the price of allowances was predicted to increase, the cost difference caused by the ETS grew significantly and became almost as important for competitiveness as energy costs by the 2030s. The scenarios further affected the total cost difference, making it both decrease and increase depending on how the ETS was altered. Depending on how the EU alters the ETS in the future, the steel sectors in some countries might become more competitive, while others might lose financial viability. Great care must therefore be taken when developing the system so that European emissions can continue to decrease without putting companies in certain countries at a severe disadvantage. Furthermore, the CBAM, or other measures, must be put in place if carbon leakage is to be avoided when the prices for ETS allowances rise. / För att motverka klimatförändringar och följa Parisavtalet, har EU introducerat ett system för utsläppshandel (ETS) och föreslagit en gränsjusteringsmekanism för koldioxid (CBAM). Även om dessa verktyg kan minska utsläppen inom EU, och i vissa fall även andra länder, så kan de påverka europeisk industri på ett oförutsägbart sätt. En industri som redan är inkluderad i ETS är stålindustrin. Hittills så har den inte blivit signifikant påverkad, eftersom priserna för utsläppsrätter har varit låga, men detta kan förändras när priserna stiger. Genom intervjuer och en litteraturstudie har faktorerna som leder till kostnadsskillnader mellan ståltillverkare i europeiska länder identifi erats och används till att skapa en modell för kostnadsskillnader. Två tillverkare med liknande produktionsmetoder har sedan jämförts för att kvantifiera kostnadsskillnaderna som beror på ETS. Möjliga förändringar till ETS har även identifierats och utvecklats till scenarier där kostnadsskillnaderna mellan de två tillverkarna som uppstått på grund av de nya förhållandena analyserats. Inköp av energi var orsaken till den absolut största kostnadsskillnaden, följt av transportkostnader då den ena tillverkaren hade en geografisk fördel i närhet till kunden. Historiskt så har kostnaden av att köp autsläppsrätter varit i princip försumbar, däremot så växer kostnadsskillnaden avsevärt när priset för utsläppsrätter ökar, och i början av 30-talet kan de vara nästan lika viktiga som inköp av energi. De identifi erade scenarierna påverkade den totala kostnadsskillnaden ytterligare. När EU vidareutvecklar ETS i framtiden kan konkurrenskraften hos nationell stålindustri både öka och minska. Utvecklingen av systemet bör därför hanteras med stor försiktighet, så att europeiska utsläpp kan fortsätta att minska utan att industrin i vissa länder får en konkurrensnackdel. För att minimera risken för koldioxidläckage när priset på utsläppsrätter stiger, är det är även viktigt att CBAM, eller liknande åtgärder, träder i kraft.

The thermal boundary condition to a fire exposed surface was quantified with a hybrid heat flux gage. Methods were developed to determine the net heat flux through the gage, incident heat flux, cold surface heat flux, convective heat transfer coefficient, adiabatic surface temperature, and the separated components of radiative and convective heat flux. Experiments were performed in a cone calorimeter with the hybrid gage flush mounted into UNIFRAX Duraboard LD ceramic board. The results were then compared to results obtained with a Schmidt-Boelter gage and a plate thermometer. The hybrid heat flux gage predicted a cold surface heat flux within 5% of cold surface heat fluxes measured with a Schmidt-Boelter gage. Adiabatic surface temperature measurements compared well with the plate thermometer measurements at steady state. Hybrid gage measurements were performed on flat plate samples of Aluminum 5083, Marinite P, and UNIFRAX Duraboard LD ceramic board. The gage and sample assemblies were exposed to mixed-mode heat transfer conditions in a cone calorimeter. Temperature measurements were performed at the top, center, bottom surfaces of the marinite and ceramic board samples. A single midpoint temperature was performed on the aluminum. Boundary condition details obtained with the hybrid gage were then input to the commercial finite element analysis package Abaqus. Abaqus was used to create the flat plate geometries of the sample and variable temperature dependent material properties were used for each material. Measured temperatures were then compared to the model predicted temperatures with good results. Hybrid gage measurements were verified using a new experimental apparatus. The apparatus consisted of an impinging jet assembly, a tungsten lamp, and a gage holster assembly. The impinging jet was used to expose the gage to isolated convection and the lamp was used to expose the gage to isolated radiation. The gage holster assembly was used to water cool the gage when desired. Measurements performed with the gage water cooled in isolated convection allowed for the convective heat transfer coefficient to be determined. Two methods were developed to determine the convective heat transfer coefficient in mixed-mode heat transfer conditions. These methods were then verified by comparison to the isolated heat transfer coefficient. Similarly, the incident radiation was isolated by water cooling the gage while only the lamp was on. The components of heat flux were then separated for mixed-mode comparisons and were verified against this isolated radiation. The hybrid gage predicted convective heat transfer coefficients within 10% of the isolated heat transfer coefficient and incident heat fluxes within 11% of the isolated radiation. / Master of Science

Les relations spatiales entre les différentes régions dans une image sont utiles pour la compréhension et l'interprétation de la scène représentée. L'analyse Spatio-temporelle d'une scène implique l'intégration du temps dans des relations spatiales entre les objets en mouvement. Les relations spatio-temporelles sont définies dans un intervalle de temps utilisant la géométrie 3D ou l'extension de la géométrie 2D à la dimension temporelle. La modélisation des relations spatiales dynamiques prend en compte la position relative des objets et leurs relations directionnelles, ceci implique les relations topologiques, directionnelles et de distance. Ces relations sont étendues au domaine temporel. Dans notre travail, on décrit une méthode de combinaison d’information topologique et directionnelle où les relations d'Allen floues 1D sont appliquées au domaine spatial. La méthode proposée intègre le flou au niveau des relations. La méthode très gourmande initialement en temps de calcul en raison de l’approximation des objets ainsi qu'à l'algorithme de fuzzification des segments des sections longitudinales est améliorée en utilisant une approximation polygonale adaptée sur les objets considérés. L'algorithme du fuzzification des segments d'une section longitudinale inclut des opérateurs d'agrégation floue. Dans la méthode proposée, Les relations topologiques 2Dsont représentées par un histogramme. Les relations floues n'étant pas exhaustives, un algorithme de défuzzification des relations spatiales a été proposé pour réaliser un ensemble JEPD de relations spatiales. Cet ensemble de relations spatiales est représenté par un graphe de voisinage où chaque nœud du graphe représente la relation topologique et directionnelle. Cette méthode définit des relations spatio-temporelles en utilisant le modèle de données Espace-Temps. Un ensemble de relations spatio-temporelles est également fourni à l'aide de la stabilité topologique. Afin de valider le modèle, nous avons développé des applications fondées sur le raisonnement spatio-temporel proposé. Celui-ci a permit la création de tables de composition pour les relations spatiales topologiques structurées en sous-tables. Les entités de ces sous-tables sont liées les unes aux autres par des relations spatiales. Dans une seconde application, nous avons proposé une méthode de prédiction des évènements entre objets en mouvement fondée sur le même raisonnement spatio-temporel. Les objets en mouvement changeant de position à chaque instant, la prédiction de la nouvelle position spatiale d'un objet tient compte des états de relations spatiales calculées précédemment. / Spatial relations between different image regions are helpful in image understanding, interpretation and computer vision applications. Spatio-temporal analysis involves the integration of spatial relations changing over time between moving objects of a dynamic scene. Spatio-temporal relations are defined for a selected time interval using 3D geometry or extension of 2D object geometry to the time dimension with sequence occurrence of primitive events for each snapshot. Modeling dynamic spatial relations takes into account the relative object position and their directional relations; this involves the topological, directional and distance relations and their logical extension to the temporal domain. In this thesis, a method for combining topological and directional relations information is discussed where 1D temporal fuzzy Allen relations are applied in spatial domain. Initially, the method has a high computational cost. This computing cost is due to the object approximation and the fuzzification algorithm of segments. The computing time has been using polygonal object approximation. Fuzzification algorithm is replaced with fuzzy aggregation operators for segments of a longitudinal section. In this method, two dimensional topological relations are represented in a histogram. The representation method for two dimensional spatial relations has been changed. These fuzzy relations are not Jointly Exhaustive and Pairwise Disjoint (JEPD). An algorithm for defuzzification of spatial relations is proposed to realize JEPD set of spatial relations, these JEPD spatial relations are represented in a neighborhood graph. In this neighborhood graph, each node represents the topological and directional relation. This method is further extended for defining spatio-temporal relations using space and time data model, a set of spatio-temporal relations are also elaborated using the stability property in topology. In an application, a method for spatio-temporal reasoning based on this new model is developed. Spatio-temporal reasoning consists of developing the composition tables for spatial relations. Composition table for topological relations are rearranged into sub-tables. Entities in these sub-tables are related to each other and mathematical rules are defined for composition of spatial relations which elaborate the relation between entities of sub-tables. In another application, we propose a method for motion event predictions between moving objects. It is a similar process to the spatio-temporal reasoning. Dynamic objects occupy different places at different time points, these objects have multiple choices for subsequent positions and a unique history. Prediction about motion events take into account the history of a moving object and predict about the semantics of a motion event.

Since the use of social media has evolved, the ability of its users to bully others has increased. One of the prevalent forms of bullying is Cyberbullying, which occurs on the social media sites such as Facebook©, WhatsApp©, and Twitter©. The past decade has witnessed a growth in cyberbullying – is a form of bullying that occurs virtually by the use of electronic devices, such as messaging, e-mail, online gaming, social media, or through images or mails sent to a mobile. This bullying is not only limited to English language and occurs in other languages. Hence, it is of the utmost importance to detect cyberbullying in multiple languages. Since current approaches to identify cyberbullying are mostly focused on English language texts, this thesis proposes a new approach (called Multilingual Cyberbullying Detection System) for the detection of cyberbullying in multiple languages (English, Hindi, and Marathi). It uses two techniques, namely, Machine Learning-based and Lexicon-based, to classify the input data as bullying or non-bullying. The aim of this research is to not only detect cyberbullying but also provide a distributed infrastructure to detect bullying. We have developed multiple prototypes (standalone, collaborative, and cloud-based) and carried out experiments with them to detect cyberbullying on different datasets from multiple languages. The outcomes of our experiments show that the machine-learning model outperforms the lexicon-based model in all the languages. In addition, the results of our experiments show that collaboration techniques can help to improve the accuracy of a poor-performing node in the system. Finally, we show that the cloud-based configurations performed better than the local configurations.

More than 1.2 million Americans are hospitalized annually with an acute coronary syndrome (ACS); many impaired quality of life after discharge with an ACS. This dissertation focuses on two novel aspects of patient health status (PHS) after ACS: how it can be predicted based on the socioeconomic status (SES) of the patient, and how it evolves over time. We used data from TRACE-CORE, a longitudinal prospective cohort of patients hospitalized with ACS. We measured PHS using both the SF-36 mental and physical component subscales (MCS and PCS) and the Seattle Angina Questionnaire (SAQ) health-related quality of life (HRQoL) and physical limitations subscales at the index hospitalization and at 1, 3, and 6-months post-discharge. Firstly, after adjusting for individual-level SES, we found that individuals living in the neighborhoods with the lowest neighborhood SES had significantly worse PHS. Secondly, we found that each of the components of PHS had subgroups with distinct patterns of evolution over time (trajectories). Both the PCS and the SAQ physical limitations subscale had two trajectories; one with average and one with impaired health status over time. For the HRQoL subscale of SAQ, we found three trajectories: Low, Average, and High scores. For MCS, we found four trajectories: High (consistently high scores), Low (consistently low scores), and two with average scores at baseline that either improved or worsened over time, referred to as Improving and Worsening, respectively. All PHS trajectories, except for MCS, predicted readmission and mortality during the 6 months to 1 year post-ACS discharge.

La modélisation de matériaux composites est un domaine d’étude qui bénéficie d’un intérêt croissant. En effet, la vulgarisation de l’utilisation de tels matériaux nécessite le développement de nouveaux modèles afin de mieux comprendre leur comportement. L’industrie automobile et aéronautique s’efforce d’optimiser le choix des matériaux en fonction des spécificités de chaque application, afin de réduire la masse des équipements et de leur assurer de meilleurs caractéristiques mécaniques et thermiques. Les matériaux composites se sont aussi présentés comme une éventuelle alternative au métal pour le rôle de blindage électromagnétique. Leur généralisation dans cette optique se heurte néanmoins à une relative méconnaissance de leur comportement électromagnétique. A cet effet, il est nécessaire de disposer de méthodologies permettant d'évaluer l'efficacité de blindage de boîtiers en matériaux composites et de cerner les différents mécanismes et paramètres correspondants.Par ailleurs, le déploiement de ces matériaux alternatifs à plus grande échelle est freiné par d'autres contraintes liées essentiellement à la difficulté de l'analyse électromagnétiques 3D complète de systèmes complexes abritant des boîtiers en matériaux composites. En effet, la complexité topologique de certains composants complique considérablement leur insertion dans les outils de simulation électromagnétique existants. De plus, le rapport d'échelle entre les différents niveaux (système, boîtiers composites, cartes, circuits, composants) est trop important ; cette disparité d'échelle complexifie considérablement la discrétisation géométrique de l'ensemble. L'association de ces différentes contraintes conduisent à des difficultés réelles aux quelles les ingénieurs CEM sont confrontés. C'est pourquoi il est nécessaire de développer des modèles performants permettant de faciliter l'analyse 3D du système hôte complet. Ce travail de thèse s'est donc réparti sur deux volets :- dans un premier temps, nous présentons une méthodologie de calcul de l'efficacité de blindage des boîtiers en matériaux composites, afin d'évaluer la potentialité de ces matériaux en termes de blindage électromagnétique et de cerner les principaux facteurs qui y contribuent.- dans un second temps et dans l'objectif de fournir une approche permettant de mettre les systèmes électroniques complexes intégrant des boîtiers de blindage composites en conformité avec les exigences strictes de CEM, nous proposons une méthodologie de modélisation des rayonnements électromagnétiques. Cette modélisation (à base d’algorithmes génétiques) permet de remplacer les dispositifs ou les boîtiers rayonnants (composites notamment) par un ensemble de dipôles élémentaires. Le modèle équivalent, de type "boîte noire", est ainsi représentatif de l’ensemble de la structure en termes de rayonnement électromagnétique en hautes fréquences et est facilement intégrable dans le maillage de structures hôtes. Ce modèle multipolaire fournit des prédictions spatiales et fréquentielles du champ électrique et magnétique permettant entre autres de calculer l'efficacité de blindage du boîtier dans l'espace, donnant ainsi un moyen de quantifier son impact perturbateur sur son environnement. D'autre part, cette approche permet de simplifier l'analyse 3D d'un système complet abritant des boîtiers composites en contrôlant le comportement EM à tous les niveaux : système, boîtiers, cartes, circuits et composants. / The modeling of composite materials is a domain of study which benefits of increasingly interest. Indeed, the popularization of the use of such materials requires the development of new models in order to better understand their behavior. The automotive and aerospace industry strives to optimize material selection based on the specificities of each application in order to reduce the weight of the equipment and to provide better mechanical and thermal characteristics. Composite materials have been also presented as a potential alternative to metals for the role of electromagnetic shielding. Their generalization in this context is nevertheless hampered by a relative lack of knowledge of their electromagnetic behavior. For this purpose, it is necessary to have methodologies to evaluate the shielding effectiveness of composite enclosures and identify the different corresponding mechanisms and parameters.Moreover, the deployment of these alternative materials on a larger scale is hindered by other constraints related mainly to the difficulty of complete 3D analysis of complex systems including composite enclosures. In fact, the topological complexity of certain components greatly complicates their integration into existing electromagnetic simulation tools. Moreover, the scale ratio between the different levels (system, composite enclosures, electronic card, circuit, component) is too large ; This disparity of scale complexifies considerably the geometrical discretization of the entire system. The combination of these different constraints leads to real difficulties to which EMC engineers face. That is why it is necessary to develop efficient models to facilitate the 3D analysis of the complete host system.This work is therefore divided in two sections :- In a first time, we present a methodology to calculate shielding effectiveness of composite enclosures of electronic equipment. The goal is to evaluate the potential of these materials in terms of electromagnetic shielding and to identify the main contributing factors.- In a second time, and in order to ensure compliance of complex electronic systems incorporating composite shielding enclosures with the stringent requirements of EMC, we propose a modeling methodology of electronic devices radiation. This modeling (based on genetic algorithms) allows to replace the radiating devices and enclosures (especially composites) by a set of elementary dipoles. The equivalent model, "black box" type, is thus representative of the entire structure in terms of high frequency electromagnetic radiation and is easily integrable in the mesh of host structures. This multipolar model provides spatial and frequency predictions of the electric and magnetic field, enabling among others to calculate the shielding effectiveness of the radiating enclosure in space, thereby giving a way to quantify its disruptive impact on its environment. Moreover, this approach allow to simplify the 3D analysis of a complete system comprising composite enclosures by controlling the EM behavior at all levels: system, enclosures, cards, circuits and components.

The emergence of Grid computing infrastructures enables researchers to shareresources and collaborate in more efficient ways than before, despite belongingto different organizations and being geographically distributed. While the Gridcomputing paradigm offers new opportunities, it also gives rise to newdifficulties. This thesis investigates methods, architectures, and algorithmsfor a range of topics in the area of Grid resource management. One studiedtopic is how to automate and improve resource selection, despite heterogeneityin Grid hardware, software, availability, ownership, and usage policies.Algorithmical difficulties for this are, e.g., characterization of jobs andresources, prediction of resource performance, and data placementconsiderations. Investigated Quality of Service aspects of resource selectioninclude how to guarantee job start and/or completion times as well as how tosynchronize multiple resources for coordinated use through coallocation.Another explored research topic is architectural considerations for frameworksthat simplify and automate submission, monitoring, and fault handling for largeamounts of jobs. This thesis also investigates suitable Grid interactionpatterns for scientific workflows, studies programming models that enable dataparallelism for such workflows, as well as analyzes how workflow compositiontools should be designed to increase flexibility and expressiveness. We today have the somewhat paradoxical situation where Grids, originally aimed tofederate resources and overcome interoperability problems between differentcomputing platforms, themselves struggle with interoperability problems causedby the wide range of interfaces, protocols, and data formats that are used indifferent environments. This thesis demonstrates how proof-of-concept softwaretools for Grid resource management can, by using (proposed) standard formatsand protocols as well as leveraging state-of-the-art principles fromservice-oriented architectures, be made independent of current Gridinfrastructures. Further interoperability contributions include an in-depthstudy that surveys issues related to the use of Grid resources in scientificworkflows. This study improves our understanding of interoperability amongscientific workflow systems by viewing this topic from three differentperspectives: model of computation, workflow language, and executionenvironment. A final contribution in this thesis is the investigation of how the design ofGrid middleware tools can adopt principles and concepts from softwareengineering in order to improve, e.g., adaptability and interoperability.