La préparation de commandes est une activité primordiale dans les entrepôts de distribution (près de 60% des coûts opérationnels dans les entrepôts traditionnels). Un des moyens de réduire ces coûts est de collecter plusieurs commandes simultanément, plutôt qu’une par une ; cela permet d’éviter des déplacements inutiles des opérateurs dans le cas de collecte manuelle des produits, et de moins solliciter les machines, dans le cas de collecte automatisée. Nous considérons dans cette thèse des entrepôts de distribution où les produits sont stockés dans des machines appelées stockeurs automatisés composés de plateaux sur lesquels les produits sont stockés ; nous prenons en compte deux types de stockeurs : les VLMs et Carousels. Ces deux types de machines, en plein développement, se différencient par le temps nécessaire pour passer d’un plateau au plateau suivant à visiter (constant dans le cas d’un VLM, dépendant du nombre d’étages entre les 2 plateaux visités dans le cadre des carousels). L’objectif de la thèse est donc de développer des méthodes pour faire des regroupements de commandes en lots dans des entrepôts automatisés afin de collecter un ensemble donné de commandes le plus efficacement possible selon des critères que nous allons définir.Nous étudions tout d’abord un premier type de regroupement de commandes en lots, pour lequel chaque lot sera collecté entièrement (toutes les commandes composant le lot seront traitées) avant de passer au suivant. Nous parlerons simplement de batching. Nous évaluons le temps de préparation de commandes, dans le cas où l’opérateur effectue la collecte sur une ou sur plusieurs machines. L’intérêt d’utiliser plusieurs machines est de permettre la recherche de plateaux en temps masqué (le stockeur effectue un changement de plateaux pendant que l’opérateur est occupé à effectuer la collecte sur d’autres stockeurs). Cette évaluation du temps de préparation de commandes nous permet d’extraire un critère d’optimisation et d’établir des modèles d’optimisation exacts pour les VLMs puis les carousels. Ces modèles sont ensuite testés avec des données réelles d’entreprise, grâce au partenariat avec l’entreprise KLS Logistic, éditeur du WMS Gildas. Enfin, nous étendons la résolution aux méthodes approchées de type métaheuristiques, afin de garantir de bonnes solutions sous un temps de calcul raisonnable. Des résultats significatifs en termes de réduction du temps de préparation de commandes permettent de justifier l’intérêt de nos travaux.Nous étudions ensuite un deuxième type de regroupement, où un ensemble de commandes est collecté simultanément mais, contrairement au cas précédent, à chaque fois qu’une commande est terminée, elle est immédiatement remplacée. Nous parlerons ici de « Rolling batching ». Cette approche est classiquement utilisée dans les « Drive ». Nous nous focalisons sur le cas de collecte effectuée sur des carousels, système le plus utilisé dans les « Drive ». Nous développons un algorithme permettant le calcul du temps d’attente de l’opérateur. Une résolution approchée couplant l’utilisation d’heuristiques et d’une métaheuristique est proposée afin de résoudre efficacement le séquencement des commandes. Nous notons que des gains significatifs sont obtenus par l’utilisation de la méthode proposée. / The order picking is a critical activity in distribution warehouses (60% of the operational costs in traditional warehouses). One way to reduce these costs is to collect several orders simultaneously, rather than individually, to avoid unnecessary travels to the operators in the case of manual picking of the products, and less workload for the machines, in the case of automated picking. We consider, in this PhD, warehouses where products are stored in automated storage systems composed by trays on which products are stored. We consider two types of machines: VLMs and Carousels. These two types differ by the time required to move from a tray to the next one to visit (constant in the case of a VLM, depending on the number of trays between 2 consecutively visited trays for the carousels). We aim to develop order batching methods in automated warehouses to collect a set of orders efficiently, according to criteria that we will define.We study a first category of order batching, for which each batch will be entirely collected (all orders associated with a given batch will be collected) before moving to the next batch. We evaluate completion time, when the operator does the picking on one or several machines. The interest of using multiple machines is to search trays in masked time (the machine change trays while the operator is busy to collect on other machines). From this evaluation of the completion time, we extract an optimization criterion and establish optimization models for the VLMs then the carousels. These models are then tested with actual data, thanks to the partnership with the company KLS Logistic, editor of the WMS Gildas. Finally, we extend the resolution to metaheuristics, to ensure good solutions under reasonable computation time. Significant savings in terms of completion time reduction justifies the interest of our work.We then study the second category, where a set of orders is collected simultaneously, but unlike the previous case, whenever an order is collected, it is immediately replaced by a new one. We'll talk here about "Rolling batching”. This approach is typically used in the "Drive-in shops". We develop an algorithm for the computation of the waiting time of the operator. Resolution coupling heuristics and a metaheuristic is proposed in order to effectively resolve the order sequencing. Significant gains are achieved by the use of the proposed method.

A computational framework for linear static aeroelastic analyses is presented. The overall aeroelasticity model is applicable to conceptual aircraft design studies and other low-fidelity aero-structural analyses. A partitioned approach is used, i. e. separate solvers for aerodynamics and structure analyses are coupled in a suitable way, together forming a model for aeroelastic simulations. Aerodynamics are modelled using the vortexlattice method (VLM), a simple computational fluid dynamics (CFD) model based on potential flow. The structure is represented by a three-dimensional (3D) Euler-Bernoulli beam model in a finite element method (FEM) formulation. A particular focus was put on the modularity and loose coupling of aforementioned models. The core of the aeroelastic framework was abstracted, such that it does not depend on any specific details of the underlying aerodynamics and structure modules. The final aeroelasticity model constitutes independent software tools for the VLM and the beam FEM, as well as a framework enabling the aeroelastic coupling. These different tools have been developed as part of this thesis work. A wind tunnel experiment with a simple wing model is presented as a validation test case. An aero-structural analysis of a fully elastic unmanned aerial vehicle (UAV) (OptiMale) is described and results are compared with an existing higherfidelity study. / Rapporten beskriver en beräkningsmodell för linjära, statisk aeroelastiska analyser. Modellen kan användas för konceptuella designstudier av flygplan. En partitionerad metod används, d v s separata lösare för aerodynamik- och strukturanalyser kopplas på ett lämpligt sätt, och bildar tillsammans en modell för aeroelastiska simulationer. Aerodynamik modelleras med hjälp av en så kallad vortex-lattice method (VLM), en enkel modell för beräkningsströmningsdynamik (CFD) som är baserad på friktionsfri strömning. Strukturen representeras av en tredimensionell (3D) Euler-Bernoulli-balkmodell implementerad med hjälp av en finita elementmetod (FEM). Ovannämnda modeller har utvecklats med fokus på modularitet och lös koppling. Kärnan i den aeroelastiska modellen har abstraherats så att den inte beror på specifika detaljer i de underliggande aerodynamik- och strukturmodulerna. Aeroelasticitetsmodellen i sin helhet består av separata mjukvaruprogram för VLM och balk-FEM, såväl som ett ramverk som möjliggör den aeroelastiska kopplingen. Dessa olika program har utvecklats som en del av examensarbetet. Ett vindtunnelförsök med en enkel vingmodell presenteras som ett valideringstest. Dessutom beskrivs en analys av ett elastiskt obemannad flygplan (OptiMale) och resultaten jämförs med en befintlig studie som har genomförts med modeller av högre trovärdighet.

Computational Fluid Dynamics has become a versatile product development tool, bringing many advantages to the conceptual design phase of products, and is being used for a wide variety of applications. In this thesis, the aerodynamics of a UAV with internal radar has been investigated. UAV is an acronym for Unmanned Aerial Vehicle, that is an aircraft which is flown with no human pilot onboard the aircraft, maneuvering the flight controls. The studies conducted throughout this thesis aimed at conducting numerical simulations of two UAV wing designs, through the utilization of the Vortex Lattice Method and Computational Fluid Dynamics. A number of different flight cases were investigated for comparing the two wing designs with respect to their aerodynamic characteristics, as well as a sustained turn analysis for the second UAV wing design. The analyses were conducted by importing provided UAV CAD-models into the numerical software, setting up appropriate grids, and running the simulations at the flight cases of interest. The results gathered from the simulations concerning the first wing design demonstrated favorable flight characteristics at lower angles of attack, although at angles of attack greater than 5°, flow separation was apparent. Regarding the results concerning the second wing design, it could be flown at greater angles of attack without experiencing flow separation, while decreasing the total drag of the UAV. For the sustained turn analysis concerning the second wing design, the aircraft generated sufficient lift at desired load factors, for maintaining its altitude however, it could not reach a trimmed state despite larger deflection angles of the ruddervators. Therefore the center of gravity position of the UAV was manipulated for sustaining the turns at desired load factors. Pressure contour plots were generated from the CFD simulations from which it could be determined that the front and nose of the aircraft could be further redesigned for decreasing aerodynamic drag.

A computational approach to analyze the hydrodynamic performance of a hydrofoil or a marine propeller undergoing unsteady motion has been developed. In order to simulate heave and pitch motion of a hydrofoil, an unsteady boundary element method based modeling is performed. The wake of the hydrofoil is modeled by a continuous dipole sheet and determined in time by applying a force-free condition on its surface. An explicit vortex core model is adapted in this model to capture the rolling up shape and to avoid instability due to roll-up deformation of the wake. The numerical results of the developed model are compared with analytical results and those from the commercial Reynolds-Averaged Navier-Stokes solver (ANSYS/FLUENT). The results show close level of agreement with each other. The problem of flow around a marine propeller performing surge, roll and heave motion in an unbounded fluid is formulated and solved using both a vortex-lattice method and a boundary element method. A fully unsteady wake alignment algorithm is implemented into the vortex-lattice method in order to satisfy the force-free condition on the propeller wake surface. Finally, a comparative study of transient propeller forces on a propeller blade obtained from BEM and VLM (with or without fully aligned wake) is carried out and results are presented. In some cases, results from the presented methods are compared with those from RANS or other numerical methods available in the literature. / text

In the digital age where video content is abundant, this thesis investigates the efficient adaptation of an existing video-language model (VLM) to new data. The research leverages CLIP, a robust language-vision model, for various video-related tasks including video retrieval. The study explores using pre-trained VLMs to extract video embeddings without the need for extensive retraining. The effectiveness of a smaller model using aggregation is compared with larger models and the application of logistic regression for few-shot learning on video embeddings is examined. The aggregation was done using both non-learning through mean-pooling and also by utilizing a transformer. The video-retrieval models were evaluated on the ActivityNet Captions dataset which contains long videos with dense descriptions while the linear probes were evaluated on ActivityNet200 a video classification dataset.  The study's findings suggest that most models improved when additional frames were employed through aggregation, leading to improved performance. A model trained with fewer frames was able to surpass those trained with two or four times more frames by instead using aggregation. The incorporation of patch dropout and the freezing of embeddings proved advantageous by enhancing performance and conserving training resources. Furthermore, using a linear probe showed that the extracted features were of high quality requiring only 2-4 samples per class to match the zero-shot performance.

This project evaluates the suitability and practicality of the program Athena Vortex Lattice (AVL) by Mark Drela. A short user guide was written to make it easier (especially for students) to get started with the program AVL. AVL was applied to calculate the induced drag and the Oswald factor. In a first task, AVL was used to calculate simple wings of different aspect ratio A and taper ratio lambda. The Oswald factor was calculated as a function f(lambda) in the same way as shown by HOERNER. Compared to HOERNER's function, the error never exceed 7.5%. Surprisingly, the function f(lambda) was not independent of aspect ratio, as could be assumed from HOERNER. Variations of f(lambda) with aspect ratio were studied and general results found. In a second task, the box wing was investigated. Box wings of different h/b ratio: 0.31, 0.62, and 0.93 were calculated in AVL. The induced drag and Oswald factor in all these cases was calculated. An equation, generally used in the literature, describes the box wing's Oswald factor with parameters k1, k2, k3 and k4. These parameters were found from results obtained with AVL by means of the Excel Solver. In this way the curve k = f(h/b) was plotted. The curve was compared with curves with various theories and experiments conducted prior by other students. The curve built based on AVL fits very well with the curve from HOERNER, PRANDTL and a second experiment made in the wind tunnel at HAW Hamburg.

No description available.

No description available.