The popularity of location empowered devices such as GPS enabled smart-phones has immensely amplified the use of location-based services in social networks. This happened by allowing users to share Geo-tagged contents such as current locations/check-ins with their social network friends. These location-aware social networks are called Location-based Social Networks (LBSN), and examples include Foursquare and Gowalla. The data of LBSNs are being used for providing different kinds of services such as the recommendation of locations, friends, activities, and media contents, and the prediction of user's locations. To provide such services, different queries are utilized that exploit activity/check-in data of users. Usually, LBSN data is divided into two parts, a social graph that encapsulates the friendships of users and an activity graph that maintains the visit history of users at locations. Such a data separation is scalable enough for processing queries that directly utilize friendship information and visit history of users. These queries are called user and activity analytic queries. The visits of users at locations create relationships between those locations. Such relationships can be built on different features such as common visitors, geographical distance, and mutual location categories between them. The process of analysing such relationships for optimizing location-based services is termed Location Analytics. In location analytics, we expose the subjective nature of locations that can further be used for applications in the domain of prediction of visitors, traffic management, route planning, and targeted marketing.In this thesis, we provide a general LBSN data model which can support storage and processing of queries required for different applications, called location analytics queries. The LBSN data model we introduce, segregates the LBSN data into three graphs: the social graph, the activity graph, and the location graph. The location graph maintains the interactions of locations among each other. We define primitive queries for each of these graphs. In order to process an advanced query, we express it as a combination of these primitive queries and process them on corresponding graphs in parallel. We further provide a distributed data processing framework called GeoSocial-GraphX (GSG). GSG implements the aforementioned LBSN data model for efficient and scalable processing of the queries. We further exploit the location graph for providing novel location analytics queries in the domain of influence maximization and visitor prediction. We introduce a notion of location influence. Such influence can capture the interactions of locations based on their visitors and can be used for propagation of information between them. The applications of such a query lie in the domain of outdoor marketing, and simulation of virus and news propagation. We also provide a unified system IMaxer that can evaluate and compare different information propagation mechanisms. We further exploit the subjective nature of locations by analysing the mobility behaviour of their visitors. We use such information to predict the individual visitors as well as the groups of visitors (cohorts) in future for those locations. The prediction of visitors can be used for better event planning, traffic management, targeted marketing, and ride-sharing services.In order to evaluate the proposed frameworks and approaches, we utilize data from four real-life LBSNs: Foursquare, Brightkite, Gowalla, and Wee Places. The detailed LBSN data mining and statistically significant experimental evaluation results show the effectiveness, efficiency, and scalability of our proposed methods. Our proposed approaches can be employed in real systems for providing life-care services. / Doctorat en Sciences de l'ingénieur et technologie / The portal is not showing my complete name. The name (my complete name), I want to have on the diploma is "Muhammad Aamir Saleem". Please correct this issue. / info:eu-repo/semantics/nonPublished

In thin films solar cells technology, many different materials have been used; however, toxic and expensive materials and techniques have been widely employed up to date. In this work, three different low/non-toxic materials were grown by means simple and cheap techniques.Cd1–xZnxS thin films with Zn concentrations of 0–5 at.% were analyzed. The effect of the addition of different molar Zn concentrations to the reaction mixture on the growth mechanism of Cd1–xZnxS thin films and the influence of these mechanisms on structural, optical and morphological properties of the films has been studied.Cd1–xZnxS thin films were synthesized by chemical bath deposition using an ammonia-free alkaline solution. Microstructural analysis by X-ray diffraction showed that all deposited films grew with hexagonal wurtzite structure and crystallite sizes decreased as the Zn concentration in the film increased. Optical measurements indicated a high optical transmission between 75 % and 90 % for wavelengths above the absorption edge. Band gap value increased from 2.48 eV to 2.62 eV, and the refractive index values for Cd1–xZnxS thin films decreased as the Zn increased. These changes in films and properties are related to a modification in growth mechanism of the Cd1–xZnxS thin films, with the influence of Zn(OH)2 formation being more important as Zn in solution increased.Characterization of ZnS thin films deposited by chemical bath in a non-toxic alkaline solution is reported. The effect of deposition technique (growth in several times) on the properties of the ZnS thin film was studied. The films exhibited a high percentage of optical transmission (greater than 80 %); as the deposition time increased a decreasing in the band gap values from 3.83 eV to 3.71 eV was observed. From chemical analysis, the presence of ZnS and Zn(OH)2 was identified and X-ray diffraction patterns exhibited a clear peak corresponding to ZnS hexagonal phase (103) plane, which was confirmed by electron diffraction patterns. From morphological studies, compact samples with well-defined particles, low roughness, homogeneous and pinhole-free in the surface were observed.On the other side, characterization of one-step electrodeposited CuInS2 thin films is reported. The effect of the complexing agent C8H5KO4 on the reduction potential of metallic ions, as well as the annealing conditions to avoid film oxidation was studied. From polarization curves, it was observed that C8H5KO4 is an appropriate complexing agent to bring the reduction potential of Cu2+ closer to In3+. XRD patterns showed films with hexagonal structure and no peaks attributed to oxide phases were observed, under a 95 % N2 + 5 % H2 atmosphere and using a carbon holder. However, others two weak peaks associated to Cu7S4 phase were identified. The CuInS2 thin films exhibited a band gap value of 1.50 eV, low optical transmission in the visible range, p-type conductivity and a charge carrier concentration of 2.3 × 10-18 cm-3. Finally, from morphological and chemical analysis, it was possible to observe a compact morphology and free of pinholes surface with a composition close to the stoichiometric one. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

The increase of greenhouse gas emission is commonly accepted to largely contribute to global warming, in part due to the massive use of non-renewable fossil energy sources. It is a reason why recently, beside other industrial sectors, electric mobility has been considered as the next generation for transportation systems. But the electrification of a vehicle introduces new challenges in its design since it involves different domains of expertise than the ones from Internal Combustion Engine (ICE) vehicles. In particular, Noise, Vibration and Harshness (NVH) comfort is significantly affected by powertrain changes, i.e. from ICE to electric machine. High and unpleasant acoustic noise from resonances may occur but can be addressed numerically to support decision making processes early enough in the design stages. The accurate prediction of its radiated acoustic noise then requires a thorough multi-physical understanding, from the system-level (electric machine) to the component-level (stator and rotor cores).First from a system-level point of view, two multi-physical modeling frameworks that use different model simplifications were implemented. By comparing simulated results to experimental measurements at each physical step of the modeling flow (electromagnetic, vibration, acoustic), it was shown that both models are accurate enough for pre-designing phases. It was also shown that considering only the stator core to contribute to the vibro-acoustic behavior of electric machines is a valid assumption.Second from a component-level point of view, the rotor and the stator core were investigated. The rotor influence on the complete machine structural dynamics was assessed. Beside the validated effects of different rotor topologies on the radiated noise, an analytical model was successfully developed to explain the occurrence of a particular vibration mode; whose explanation was still not offered in literature. In parallel, the stator core was studied, essentially because it is composed of hundreds of thin laminations stacked together which introduce difficulties in understanding its structural behavior. The effects of the laminations on the structural behavior of the stator core were studied numerically and experimentally as well. Two modeling guidelines were thus provided depending on the mode shape of interest and the computational resources available. The experimental studies comforted these two modeling approaches, and also permitted to highlight the importance of looking at the damping properties. Therefore it was shown that different lamination stacking techniques could affect significantly this damping.Finally the influence of the stacking technique (gluing, welding) on the structural behavior of the laminated compound motivated the implementation of an alternative solution to the mitigation of resonance phenomenon responsible for large acoustic noise. By using a skewed distribution of welding or glue lines, the technique aims at forcing laminations to vibrate with different phases which generates friction between them. The induced damping increases and then depends on the introduced asymmetry and on the mode shape considered. This innovative technique was validated experimentally and showed up to 7 times higher structural damping and 10 dB reduction in structural transfer function amplitudes. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

The consequences of over-reliance on fossil fuels for energy supply, namely climate change and security of supply, are pushing for the use of local, renewable energy sources which are usually variable in nature, prompting the need for energy storage. Today, there is a trend towards distributed energy storage, justified by the distributed nature of renewable energy sources and the important share of energy consumption in buildings. Important information on such small scale energy storage installations, however, is still missing and the results of the existing literature vary widely. To account for these research gaps, a thorough characterisation of energy storage technologies is performed, together with the dimensioning and optimisation of such installations in buildings, as well as some aspects of their impact on the grid.It is found that storage is still far from grid parity and expensive when compared to other solutions, although necessary for a high share of renewables. Also, energy storage is subject to important economies of scale and technical limitations that counter the reasoning for a distributed approach. There is an important lack of practical information on several energy storage technologies, and many studies on distributed storage use downsized values from large-scale installations that do not correctly depict smaller installations, leading to biased results. Nevertheless, today, lithium-ion batteries seem to be the most appropriate electrical energy storage technology for buildings, being well adapted to short term storage. On the other hand, a very high share of renewables will push for long term storage, itself a challenge given the high cost brought by a low utilisation factor. A high share of distributed generation also impacts the grid, a problem which most final consumers have no economic incentive to mitigate. Storage by itself, without a sound control strategy, does not help as it tends to increase the load variability while the peak load remains the same. Specific control algorithms could change that but incentives must be present, namely through the adaptation of current grid tariffs that do not correctly allocate existing costs. These findings are essential in the future planning of energy systems as well as in energy policy. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

With the recent growth of social media platforms and the human desire to interact with the digital world a lot of human-human and human-device interaction data is getting generated every second. With the boom of the Internet of Things (IoT) devices, a lot of device-device interactions are also now on the rise. All these interactions are nothing but a representation of how the underlying network is connecting different entities over time. These interactions when modeled as an interaction network presents a lot of unique opportunities to uncover interesting patterns and to understand the dynamics of the network. Understanding the dynamics of the network is very important because it encapsulates the way we communicate, socialize, consume information and get influenced. To this end, in this PhD thesis, we focus on analyzing an interaction network to understand how the underlying network is being used. We define interaction network as a sequence of time-stamped interactions E over edges of a static graph G=(V, E). Interaction networks can be used to model many real-world networks for example, in a social network or a communication network, each interaction over an edge represents an interaction between two users, e.g. emailing, making a call, re-tweeting, or in case of the financial network an interaction between two accounts to represent a transaction.We analyze interaction network under two settings. In the first setting, we study interaction network under a sliding window model. We assume a node could pass information to other nodes if they are connected to them using edges present in a time window. In this model, we study how the importance or centrality of a node evolves over time. In the second setting, we put additional constraints on how information flows between nodes. We assume a node could pass information to other nodes only if there is a temporal path between them. To restrict the length of the temporal paths we consider a time window in this approach as well. We apply this model to solve the time-constrained influence maximization problem. By analyzing the interaction network data under our model we find the top-k most influential nodes. We test our model both on human-human interaction using social network data as well as on location-location interaction using location-based social network(LBSNs) data. In the same setting, we also mine temporal cyclic paths to understand the communication patterns in a network. Temporal cycles have many applications and appear naturally in communication networks where one person posts a message and after a while reacts to a thread of reactions from peers on the post. In financial networks, on the other hand, the presence of a temporal cycle could be indicative of certain types of fraud. We provide efficient algorithms for all our analysis and test their efficiency and effectiveness on real-world data.Finally, given that many of the algorithms we study have huge computational demands, we also studied distributed graph processing algorithms. An important aspect of these algorithms is to correctly partition the graph data between different machines. A lot of research has been done on efficient graph partitioning strategies but there is no one good partitioning strategy for all kind of graphs and algorithms. Choosing the best partitioning strategy is nontrivial and is mostly a trial and error exercise. To address this problem we provide a cost model based approach to give a better understanding of how a given partitioning strategy is performing for a given graph and algorithm. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

3D reinforced composites are favored for aerospace, automotive and wind turbine applicationbecause of their high specific stiffness and strength in the in-plane and out-ofplanedirections. In these composites, pins, stitching yarns and binder yarns are insertedthrough-the-thickness of the in-plane fiber-reinforced regions. Binder parameters as diameter,content, pattern and tensioning can further be varied to regulate the out-of-planeproperties. However, the insertion of these binders distorts the reinforcement which furthercan affect the global and local mechanical behaviour. Unit-cell models offered avaluable approach to assess the effect of the distortions on these mechanical features.An approach is presented to include the main geometrical features of pinned, stitchedand 3D woven composites into mesoscopic unit-cell models. Discretised lines, whichrepresent the main geometrical features, are hereby gradually shaped by geometrical operationswhile a geometrical contact treatment account for line interactions. The localfiber volume fraction and fiber direction distributions are afterwards modelled on crosssectionsin a post-processing step. Tools are further proposed to automatically transformthe geometrical models into finite element models. The effect of distortions, local fibervolume fraction and fiber direction, and typical geometrical features for each 3D reinforcedcomposite, on the stiffness and damage initiation stress levels is investigated bymeans of elastic finite element (FE)-computations.The shape of geometrical features corresponding to the different binder parameters couldautomatically be generated and the dimensions of features could be controlled by the parametersof the geometrical operations. The stiffness of a 3D reinforced composite havebeen observed to be either decreased or increased (dependent on the stacking sequence,the binder type and the loading direction). Early damage initiation in the FE-modelswas observed to take place near the binder locations, which was mainly caused by transverseand shear cracking in the fiber-reinforced regions. Local fiber volume fraction andfiber direction have shown to affect damage initation mechanisms and stress levels, andshould therefore be properly included in the models. In future work, the possibility ofthe framework to generate unit-cells including voids and micro-vascular networks canbe investigated and the finite element models can be extended with damage and crackpropagation mechanisms for damage and failure computations. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

High-mast light poles are used frequently to illuminate large areas such as motorways and parking lots. These poles are extremely tall with respect to their cross-section, reaching heights of more than 40 meters. These structures undergo a strong aeroelastic response due to wind, oftentimes resulting in fatigue cracking at the base. The purpose of the research is to better understand the effects of wind-induced vibrations of tall flexible structures using a combination of computational fluid dynamics and structural finite element codes. Field results of existing high-mast poles will be used to calibrate and verify the theoretical modeling.Periodic vortex shedding is observed to occur on these structures at certain wind velocities. The shedding of vortices causes pressure differences across the pole, resulting in a net driving force perpendicular to the direction of the wind. When the frequency of shedding, and thus the driving force, matches the natural frequency of the pole, excitation of the structure can be significant. This phenomenon is called lock-in. Poles that are repeatedly subjected to wind at lock-in velocity may suffer excessive deformation and fatigue damage. The aeroelastic response is especially significant, since the damping of the structural system is so small.In order to model the fluid-structure interaction, OpenFOAM libraries were compiled into a single application that combined a structural dynamic finite element code along with a mesh movement algorithm. The loosely coupled system applies the driving forces (integrated pressures) to the structure in a conventional serial staggered procedure. The coupling of the two domains and the mesh deformation calculationswere software written by the author. The 3-field solution formulation is implemented using a mesh movement algorithm based on a pseudo-elastic approach. Incompressible flow is assumed, as the lock-in velocities for the first three natural frequency modes ofthe pole are relatively low. Large Eddy Simulation is used for turbulence modeling.In conjunction with the University of Wyoming, two existing steel hexadecagonal high-mast poles in Wyoming, USA, were instrumented with accelerometers and anemometers. These data were used to calibrate and verify the structural, stiffness, damping, and response characteristics.A series of 14 simulations were run that increased in the difficulty of the domain being simulated. Different aspects of the pole aparatus were investigated individually, such as the taper and angle of incidence of flow. An atmospheric boundary layer model was incorporated. The final case resulted in the simulation of a 16-sided tapered pole subject to flow from an atmospheric boundary layer inlet, incorporating large eddy simulation turbulence modeling. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished

Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished

Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished