Dans ce travail de thèse, nous analysons la structure crustale de la marge est-algérienne et du bassin adjacent (région d’Annaba), à partir d’un ensemble de nouvelles données acquises durant la Campagne SPIRAL’2009 incluant un profil sismique terre-mer de ~240 km de long, des lignes sismiques réflexion pénétrante 360-traces, et des profils gravimétriques et magnétiques. Nous avons par ailleurs disposé pour cette étude de données complémentaires incluant notamment un ensemble de profils de sismique réflexion offrant des résolutions complémentaires. La structure crustale ainsi établie nous permet de discuter les nombreux modèles cinématiques d’ouverture du bassin est-algérien proposés dans la littérature, afin de caler dans le temps la formation du bassin par rapport à la collision. Elle permet également de discuter la localisation de la déformation liée à la réactivation de la marge, par rapport aux grands domaines lithosphériques du système marge-bassin, afin de mieux comprendre les modalités de l’inversion. Dans le bassin profond, la modélisation directe des temps d’arrivée et des amplitudes des ondes réfractées et réfléchies met en évidence une croûte océanique anormalement mince de 5-5.5 km d’épaisseur, composée de deux couches. La première, de 2.2 km d’épaisseur, montre des vitesses comprises entre 4.8 à 6.0 km/s impliquant un fort gradient; la seconde de 3.3 km d’épaisseur, présente des vitesses comprises entre 6.0 à 7.1 km/s et un plus faible gradient de vitesse. La modélisation des temps d’arrivées des ondes S fourni pour cette couche un coefficient de Poisson de 0.28, indiquant qu’elle est majoritairement constituée de gabbros. / In this study, we determine the deep structure of the eastern Algerian basin and its southern margin in the Annaba region (easternmost Algeria), to better constrain the plate kinematic reconstruction in this region. This study is based on new geophysical data collected during the SPIRAL cruise in 2009 that included a wide-angle, 240-km-long, onshore-offshore seismic profile, multichannel seismic reflection lines, and gravity and magnetic data, which was complemented by the available geophysical data for the study area. The analysis and modeling of the wide-angle seismic data using travel-times and amplitudes, and integrated with the multichannel seismic lines, reveal the detailed structure of an ocean-to-continent transition. In the deep basin, there is an ~5.5-km-thick oceanic crust that is composed of two layers. The upper layer of the crust is defined by a high velocity gradient and P-wave velocities between 4.8 km/s and 6.0 km/s from the top to the bottom. The lower crust is defined by a lower velocity gradient and P-wave velocity between 6.0 km/s and 7.1 km/s. The Poisson ratio in the lower crust deduced from S-wave modeling is 0.28, which indicates that the lower crust is composed mainly of gabbros. Below the continental edge, a typical continental crust with P-wave velocities between 5.2 km/s and 7.0 km/s from the top to the bottom shows a gradual seaward thinning of ~15 km over an ~35-km distance.

This paper is focus on flowers recognition in an image and class classification. Theoretical part is focus on problematics of deep convolutional neural networks. The practical part if focuse on created flowers database, with which it is further worked on. The database conteins it total 13000 plant pictures of 26 spicies as cornflower, violet, gerbera, cha- momile, cornflower, liverwort, hawkweed, clover, carnation, lily of the valley, marguerite daisy, pansy, poppy, marigold, daffodil, dandelion, teasel, forget-me-not, rose, anemone, daisy, sunflower, snowdrop, ragwort, tulip and celandine. Next is in the paper described used neural network model Inception v3 for class classification. The resulting accuracy has been achieved 92%.

This thesis deals with designing an automatic generator of deep neural networks for image classification. Theoretical part clarifies what a neural network and formal neuron are. Furthermore, the types of neural network architectures are presented. The focus of this thesis is convolutional neural networks, several pieces of research from this field are mentioned. The practical part of this thesis describes information with regards to the implementation of neural network generator, possible frameworks and programming languages for such implementation. Brief description of the implementation itself is presented as well as implemented layers. Generated neural networks are tested on Google-Landmarks dataset and results are commented upon.

AMI Meeting Corpus (AMI) -databasen används för att undersöka igenkännande av gruppaktivitet. AMI Meeting Corpus (AMI) -databasen ger forskare fjärrstyrda möten och naturliga möten i en kontorsmiljö; mötescenario i ett fyra personers stort kontorsrum. För attuppnågruppaktivitetsigenkänninganvändesbildsekvenserfrånvideosoch2-dimensionella audiospektrogram från AMI-databasen. Bildsekvenserna är RGB-färgade bilder och ljudspektrogram har en färgkanal. Bildsekvenserna producerades i batcher så att temporala funktioner kunde utvärderas tillsammans med ljudspektrogrammen. Det har visats att inkludering av temporala funktioner både under modellträning och sedan förutsäga beteende hos en aktivitet ökar valideringsnoggrannheten jämfört med modeller som endast använder rumsfunktioner[1]. Deep learning arkitekturer har implementerats för att känna igen olika mänskliga aktiviteter i AMI-kontorsmiljön med hjälp av extraherade data från the AMI-databas.Neurala nätverks modellerna byggdes med hjälp av KerasAPI tillsammans med TensorFlow biblioteket. Det finns olika typer av neurala nätverksarkitekturer. Arkitekturerna som undersöktes i detta projektet var Residual Neural Network, Visual GeometryGroup 16, Inception V3 och RCNN (LSTM). ImageNet-vikter har använts för att initialisera vikterna för Neurala nätverk basmodeller. ImageNet-vikterna tillhandahålls av Keras API och är optimerade för varje basmodell [2]. Basmodellerna använder ImageNet-vikter när de extraherar funktioner från inmatningsdata. Funktionsextraktionen med hjälp av ImageNet-vikter eller slumpmässiga vikter tillsammans med basmodellerna visade lovande resultat. Både Deep Learning användningen av täta skikt och LSTM spatio-temporala sekvens predikering implementerades framgångsrikt. / The Augmented Multi-party Interaction(AMI) Meeting Corpus database is used to investigate group activity recognition in an office environment. The AMI Meeting Corpus database provides researchers with remote controlled meetings and natural meetings in an office environment; meeting scenario in a four person sized office room. To achieve the group activity recognition video frames and 2-dimensional audio spectrograms were extracted from the AMI database. The video frames were RGB colored images and audio spectrograms had one color channel. The video frames were produced in batches so that temporal features could be evaluated together with the audio spectrogrames. It has been shown that including temporal features both during model training and then predicting the behavior of an activity increases the validation accuracy compared to models that only use spatial features [1]. Deep learning architectures have been implemented to recognize different human activities in the AMI office environment using the extracted data from the AMI database.The Neural Network models were built using the Keras API together with TensorFlow library. There are different types of Neural Network architectures. The architecture types that were investigated in this project were Residual Neural Network, Visual Geometry Group 16, Inception V3 and RCNN(Recurrent Neural Network). ImageNet weights have been used to initialize the weights for the Neural Network base models. ImageNet weights were provided by Keras API and was optimized for each base model[2]. The base models uses ImageNet weights when extracting features from the input data.The feature extraction using ImageNet weights or random weights together with the base models showed promising results. Both the Deep Learning using dense layers and the LSTM spatio-temporal sequence prediction were implemented successfully.

Turbulent and cavitation induced pipe vibration is a large problem in industry often resulting in pipe failures. This thesis provides an experimental investigation on turbulent flow and cavitation induced pipe vibration caused by sharp edged baffle plates. Due to large pressure losses across a baffle plate, cavitation can result. Cavitation can be destructive to pipe flow in the form of induced pipe wall vibration and cavitation inception. Incipient and critical cavitation numbers are design points that are often used in designing baffle plate type geometries. This investigation presents how these design limits vary with the influencing parameters by exploring a range of different baffle plate geometries. The baffle plates explored contained varying hole sizes that ranged from 0.159 cm to 2.54 cm, with the total through area, or openness, of each baffle plate ranging between 11% and 60%. Plate thickness varied from 0.32–0.635 cm. Reynolds numbers ranged from 5 x 10^4 -85 x 104. The results show that the cavitation design limits are function of size scale effects and the loss coefficient only. The results also show that the loss coefficient for a baffle plate varies not only with total through area ratio, but also due with the plate thickness to baffle hole diameter ratio. Pipe wall vibrations were shown to decrease with increased through area ratio and increased thickness to diameter ratios. An investigation was also performed to characterize the attenuation of vibration in the streamwise direction of a baffle plate. It was show that the attenuation was largely effected by the presence of cavitation. Attenuation was shown to be a function of the geometry of the baffle plate. This work resulted in empirical models that can be used for predicting pipe vibration levels, the point of cavitation inception, and the streamwise distance where the attenuation of vibration levels caused by a baffle plate occurs.

The global use of solar power is growing at an unprecedented rate, making the need toaccurately track the energy generation of decentralised solar energy systems (SES) more andmore relevant. The purpose of this thesis is to further develop a binary image classifier for thesimulation system framework known as Alfrödull, which will be used to detect and segment SESfrom aerial images to simulate the energy generation within a given Swedish municipality on anhourly basis. This project focuses on improving the Alfrödull classifier through four differentanalyses. the first focusing on examining how additional training data from publicly availabledatasets affects the model performance. The second on how the model can be improvedthrough the use of various image pre-processing techniques. The third on how the model canbe improved through balancing the training datasets to make up for the low amount of positiveimages as well as utilising model ensembles for joint classification. Finally, the fourth analysisemploys a sliding window approach to classify overlapping image tiles. The results show thathaving training data that is a good representation of the environment the model will be used in iscrucial, that the use of image augmentation policies can significantly improve modelperformance, that compensating for class imbalance as well as utilising ensemble methodspositively impacts model performance and that a sliding window approach to classifyingoverlapping images significantly decreases the amount of missed SES at the cost of clusters offalsely classified negative images (false positives). In conclusion, this thesis serves as animportant stepping stone in the practical implementation of the Alfrödull framework, showcasingthe key aspects in making a well performing binary image classifier of SES in Sweden.

The developments of new types of conductors and increase of voltage level have driven the need to carry out research on evaluating overhead line acoustic noise. The surface potential gradient of a conductor is a critical design parameter for planning overhead lines, as it determines the level of corona loss (CL), radio interference (RI), and audible noise (AN). The majority of existing models for surface gradient calculation are based on analytical methods which restrict their application in simulating complex surface geometries. This thesis proposes a novel method which utilizes both analytical and numerical procedures to predict the surface gradient. Stranding shape, proximity of tower, protrusions and bundle arrangements are considered within this model. One of UK National Grid's transmission line configurations has been selected as an example to compare the results for different methods. The different stranding shapes are a key variable in determining dry surface fields. The dynamic behaviour of water droplets subject to AC electric fields is investigated by experiment and finite element modelling. The motion of a water droplet is considered on the surface of a metallic sphere. To understand the consequences of vibration, the FEA model is introduced to study the dynamics of a single droplet in terms of phase shift between vibration and exciting voltage. Moreover, the evolution of electric field within the whole cycle of vibration is investigated. The profile of the electric field and the characteristics of mechanical vibration are evaluated. Surprisingly the phase shift between these characteristics results in the maximum field occurring when the droplet is in a flattened profile rather than when it is ‘pointed’.Research work on audible noise emitted from overhead line conductors is reviewed, and a unique experimental set up employing a semi-anechoic chamber and corona cage is described. Acoustically, this facility isolates undesirable background noise and provides a free-field test space inside the anechoic chamber. Electrically, the corona cage simulates a 3 m section of 400 kV overhead line conductors by achieving the equivalent surface gradient. UV imaging, acoustic measurements and a partial discharge detection system are employed as instrumentation. The acoustic and electrical performance is demonstrated through a series of experiments. Results are discussed, and the mechanisms for acoustic noise are considered. A strategy for evaluating the noise emission level for overhead line conductors is developed. Comments are made on predicting acoustic noise from overhead lines. The technical achievements of this thesis are summarized in three aspects. First of all, an FEA model is developed to calculate the surface electric field for overhead line conductors and this has been demonstrated as an efficient tool for power utilities in computing surface electric field especially for dry condition. The second achievement is the droplet vibration study which describes the droplets' behaviour under rain conditions, such as the phase shift between the voltage and the vibration magnitude, the ejection phenomena and the electric field enhancement due to the shape change of droplets. The third contribution is the development of a standardized procedure in assessing noise emission level and the characteristics of noise emissions for various types of existing conductors in National Grid.

Goal of my work is to introduce software development process metods specialized to Rational Unified Process metod from IBM. I aplicate inception and elaboration phases of RUP on practical example.

IBM Rational Unified Process is a robust iterative software development methodology. The main goal of master´s thesis is to describe the aspects of this methodology and to analyze and design a fictional information system within the scope of "inception" and "elaboration" phase.

The thesis deals with a problem of the design of information system. The design of system is provided with Rational Unifeid Process methodology. This thesis creates a list of requests to system. It makes an analysis and a design of current information system. It useses a RUP methodology to realize two first phases, inception phase and elaboration phase. Created elaboration phase of document is a base for creating programming prototype in Ruby on Rails environment using Ruby language with combination of HTML code.