APRÈN. Portal de la producció docent del professorat de la UPC

This project aims to study the structure of the Airbus A380-800, the largest commercial aircraft ever built, and to create a 1:50 scale model of its structure. First, an exhaustive investigation is conducted on the structural configuration of the A380, gathering the blueprints and specifications of the fuselage, wings, and stabilizers. Then, all the three-dimensional parts are modelled using SolidWorks, focusing on maintaining the aircraft's real structure. Once the 3D design is completed, the physical model's parts are manufactured. Laser cutting is used on DM board for most components, while 3D printing is employed for more complex parts, such as the wing tips. At this stage, material selection and techniques are key to ensuring the model's strength and accuracy. Finally, all the parts are assembled, resulting in a model that faithfully represents the structure of the Airbus A380. The project concludes with several recommendations for future studies, such as the possibility of adding more complex components, conducting a detailed structural analysis, or expanding the model with additional elements like flight control systems.

This thesis aims to define the business model of Halcyon, Hermeus' commercial passenger aircraft, to analyze its viability. This is achieved using the Business Model Canvas, a tool that allows visualizing the business model studied. To accomplish this, the origin and current market of hypersonic passenger aircraft must first be understood through a state of the art review. Following this, a Case Study is conducted to deeply examine the most relevant aspects of the company. Once the Business Model Canvas is developed, a risk analysis is performed to identify the weaknesses and strengths of the model and how to address potential complications. Finally, the project includes an environmental impact study to determine if the technology is sustainable and what key players are doing to address this issue.

Computer Vision tasks, like Object Detection, could become essential tools for traffic management in the near future. However, current object detection models need considerable improvements to be deployable in real-world applications. The aim of this thesis is to train a traffic object detection model specialized in drone footage, and document the process to provide a framework for future research and development. The trained model has to be capable of detecting multiple vehicle classes in real-time video feeds. Using YOLOv8 and a customized version of the VisDrone dataset that merges similar object classes, such as 'car' and 'van', a model was trained that achieved the following performance metrics: mAP50 0.651, mAP50-95 0.44, 25.57 FPS for the image speed test, and 31.71 FPS for the video speed test. This model can not be considered ready for deployment in real-world applications yet, but a clear path for improving it has been defined in this report. The reduced number of classes, compared to other models trained with default VisDrone, makes this model more practical than several other models trained by other researchers. In addition to this custom dataset model, seven other models were trained through the length of this project. Among them, there is a model trained with the regular VisDrone dataset that achieved the following metrics: mAP50 0.551, mAP50-95 0.353, 21.32 FPS for the image speed test and 23.33 FPS for the video speed test.

El siguiente trabajo aborda la convergencia entre el diseño y la ingeniería a través del desarrollo de unos nuevos soportes para el alerón de un Porsche 992 GT3 RS. En el departamento de 'Studio Engineering' de Porsche, donde se actúa como intermediario entre la creatividad del diseño y la rigurosidad técnica, surge la necesidad de crear soluciones que armonicen la estética con el rendimiento exigido en la marca. Este trabajo aparece como una puesta en práctica de este proceso. Se realizarán diversos diseños preliminares del concepto, que se irán depurando en función de los requerimientos de diseño y aerodinámica. Finalmente, se harán simulaciones CFD con las que se analizará la viabilidad técnica del alerón. Se espera que los resultados obtenidos no solo contribuyan al desarrollo técnico, sino que también sirvan como un ejemplo de integración exitosa entre ingeniería y diseño en la industria automotriz de alto rendimiento.

The thesis focuses on Exosonic's creation of low-boom supersonic jets and examines their business strategy and market tactics. This research explores Exosonic's whole business strategy, looking at important elements such their value proposition, income streams, client groups, and cost structure. It draws attention to how Exosonic occupies a distinct position in the aerospace sector, offering its cutting-edge supersonic aircraft options to both wealthy private purchasers and commercial airlines. In order to highlight their crucial roles in Exosonic's operational performance and market penetration plans, key collaborations with aerospace supply chain partners, technology collaborators including research institutes and NASA, as well as strategic alliances in marketing and worldwide services, are studied. The thesis also evaluates the prospects and obstacles facing the supersonic travel sector, such as legal restrictions, environmental issues, and the technical developments that allow Exosonic to provide quieter, faster aircraft. A close examination of customer interactions and connections reveals how Exosonic establishes and preserves relationships via committed sales teams, expert chartering services, and extensive post-purchase assistance. The goal of the thesis is to present a comprehensive analysis of Exosonic's strategic approach to reinventing supersonic flight and making it appealing to modern aviation markets. This study gives useful information for industry players interested in the next wave of air transport, in addition to contributing to scholarly discussions on supersonic aviation business strategies. In a short resume, a case study and a business model analysis are done.

The project's aim is to conduct a preliminary computational study on composite fibre materials for aerospace applications. To achieve this, the root problem will be addressed by investigating the possibility of simulating laminates made with solids instead of shells to obtain small-scale results, offering insights into their performance on a larger scale. To achieve this objective, a state-of-the-art review will contextualize the current situation of FEM, the use of FRP in the aerospace field, and of the software used, Inventor Nastran. Next, the theoretical framework will provide all the necessary equations and information to understand and carry out the simulations (stiffness matrix calculation, how meshing works, etc.). This will be followed by a review of the applicable regulations for a UAV, as it is the type of aircraft manufactured by the company. Next, the materials that will be used for the calculations will be presented, as are the ones used by Singular Aircraft, and their properties will be calculated. The laminate simulations will then be carried out following the FEM. Primarily, two simulations will be conducted: a simple tensile analysis, where one face of the laminate is fixed and a distributed force is applied to the opposite, and a four-point bending test analysis, where two constraints and two vertical forces are added to the laminate. During these simulations, certain notable problems and errors are encountered, such as the software's inability to recognize local coordinates. Finally, a budget of the work will be prepared, an analysis and evaluation of the project's environmental and social implications will be conducted, and to sum up, some brief conclusions will be drawn.

This work presents the meticulous construction of a 1:12 scale model of the A-10 Thunderbolt II aircraft, known for its robustness and close support mission capability. The model reproduces the most distinctive features of the A-10 Thunderbolt II, a legendary aircraft for its unique design, operational history, survivability and strike capabilities, and battlefield performance. In order to carry out the work, a research of its development, history, upgrades and improvements, mission history and the beginning of its retirement was carried out. Then a study was made of the parts of the aircraft in order to make a 1:12 scale model of its structure. Once the modelling was completed, the plans were drawn up to obtain the parts that were later used to assemble the model. The model has been built with poplar wood, as this is light and quite flexible, characteristics that make it ideal for modelling. The final result of the model is faithful to the structure of the aircraft and at the same time it serves as an informative element about aerospace engineering. An economic study has also been carried out on the costs of the project, as well as a study on its environmental and social impact, studying the CO2 emissions produced during the duration of the project and analysing the sustainable development objectives created by the UN and applied to the project.

El treball consisteix a dur a terme una anàlisi estructural del Boeing 747-8 Intercontinental, fer el disseny tridimensional de l'estructura i construir el model a escala 1:50. L'elecció d'aquest model es troba justificada per la peculiar forma que presenta el seu fuselatge i per la importància que ha tingut a la història de l'aviació. Inicialment, s'ha realitzat una recerca històrica sobre l'aeronau i s'ha explicat els detalls de les diferents versions que s'han construït al llarg dels anys fins a arribar a l'última i més recent, la -8i. Posteriorment, s'ha realitzat un estudi sobre la geometria i l'estructura que conformen aquest model i seguidament s'ha procedit a dissenyar amb dibuix assistit per ordinador (CAD) cadascuna de les parts de l'aeronau. S'han respectat les dimensions i geometries reals buscant sempre la màxima precisió i semblança possible amb l'avió real. No obstant això, han estat necessàries algunes modificacions per a poder adaptar correctament el disseny a escala 1:50. Finalment, s'ha construït el model a escala amb les peces prèviament dissenyades. El resultat obtingut ha estat un model 3D que s'assimila en gran part al model real i compleix amb les expectatives marcades a l'inici del treball.

The field of computational fluid dynamics (CFD) is concerned with the research and resolution of a wide variety of problems related to transport phenomena, from the study of the flow around a flying vehicle to the simulation of industrial heating, ventilating and air conditioning (HVAC), or the investigation of meteorology and climate change. The significance of progress in the simulation algorithms and methods for fluid mechanics and heat transfer can be observed in many engineering and scientific applications. The objective of this final bachelor thesis is the computational resolution of the Navier-Stokes equations for incompressible flows. After a deep review in the mathematical description of the physical phenomena, which leads to the conservation equations, and the numerical methods most commonly found in CFD, which leads to the discretization of the equations, a series of C++ codes have been developed in order to solve several problems related to fluid mechanics and heat transfer via the finite volume method (FVM). The results obtained from the codes are then thoroughly verified with benchmark data provided by reference scientific articles and by the Centre Tecnològic de Transferència de Calor (CTTC) from the Universitat Politècnica de Catalunya (UPC). First of all, a simple transient heat conduction case in two dimensions is solved in order to evaluate the temporal evolution of the temperature field inside a solid domain. Then, the discretization of the general convection-diffusion equation is carried out with the introduction of numerical schemes that allow to approximate the nonlinear convective terms in the equation, and the procedure is applied to various cases with a given flow field. After that, the fractional step method (FSM) is presented in order to solve the Navier-Stokes equations for incompressible flows and three cases involving different phenomena, such as natural convection and external flow, are solved with this method. Finally, an introduction to turbulent flows and turbulence modeling is also included and tested with the resolution of the Burgers equation in Fourier space.

This study investigates the boundary layer of a high-pressure compressor blade in a gas turbine by applying five different methodologies, including three newly developed ones based on turbulence intensity. The methodologies were developed and employed to analyse high-fidelity data from wall-resolving large eddy simulations (WRLES). The objective was to compare the methods, understand their strengths and weaknesses, and select the most accurate method for characterising the boundary layer. After applying the five methods to different simulation cases, their results were compared and analysed. It was found that the Diagnostic method, which combines the Gradient method in laminar regions with a turbulence-based method proposed by Vinuesa et al., provided the most satisfactory, consistent, and accurate results for boundary layer analysis in the studied cases. However, to improve the method and make it capable of addressing a wider range of scenarios, it is suggested to investigate and optimise the regime identification parameter due to inefficiencies observed in the results.

The present study analyzes the implementation of a novel framework to tackle the computational costs and complexity associated with performing the inverse problem that rules the design of functional metamaterials and that, nowadays, impedes the fast incorporation of the latter into most markets. Specifically, a novel framework featuring a variational autoencoder and a regressor trained with a substantial database is developed with the aim of generating a low-dimensional, continuous, and organized latent space that enables the use of vectorial operations to transform the encoded microstructures. Taking this as the base of the study, different procedures are introduced to generate microstructures' families and multiscale systems, with the final goal of setting the foundation for the strategy developed to solve the multiscale design of functional metamaterials without the need for iterative procedures. Conjointly, the different stages of the present study and the generated latent space prove the feasibility of the proposed framework and the considerable reduction in computational costs. Additionally, a brief insight into the wide range of applications that an efficient generation of functional metamaterials has is included in the present study. However, because of the great complexity associated with the proper generation and organisation of the latent space, as well as the deep understanding of the latter, a study prior to the development of the proposed framework is carried out. This prior examination, mainly involves the progressive coding of a variational autoencoder without the use of native functions to guarantee a more detailed comprehension of the different machine learning models, as well as different analyses of the obtained results. More importantly, it is done by replicating the original framework, but for a more simple problem to ensure a lower computational expense and, therefore, enable the performance of more efficient tests and experiments that serve to gain expertise in the matter.

Debido al continuo crecimiento de los accidentes en las carreteras, se ha realizado el estudio del diseño de un dron capaz de medir la velocidad de los vehículos y leer sus matrículas. Esto supone una mejora respecto de los radares de velocidad actuales, ya que es mucho más barato que cualquier sistema actual, es insonoro respecto de los vehículos y, además, puede llegar a ser invisible para los conductores. Para desarrollar este proyecto, primero se han elegido todos los componentes necesarios como: controladora de vuelo, sensor de movimiento, cámara y ordenador a bordo. Más adelante, se hace la programación para que el sistema funcione de la siguiente manera: el dron se desplaza automáticamente mediante coordenadas que previamente se le han indicado, el sensor de movimiento comenzará a funcionar una vez el dron este estático y si el sensor detecta una velocidad por encima del límite, la cámara realizará una foto. Esta fotografía más adelante será procesada utilizando CNNs para leer los dígitos de las matrículas. Una vez acabada la programación, se realizan las pruebas pertinentes para comprobar que el sistema integrado funcione adecuadamente. En dichas pruebas, el dron demostró que es capaz de medir la velocidad de los objetos, capturar imágenes y leer los dígitos de las matrículas, cumpliendo así el objetivo principal del proyecto. En cuanto a los resultados, el dron de vigilancia desarrollado ofrece una solución viable y eficiente para la vigilancia del tráfico. Además, se concluye que este sistema no solo mejora la seguridad vial, sino que también reduce los costos y el impacto ambiental en comparación con los métodos tradicionales de control de velocidad.

Este trabajo final de grado se centra en implementar un sistema de bajo coste para el control de un dron autónomo destinado a realizar inspecciones visuales de cualquier aeronave comercial. La motivación de este proyecto surge de la necesidad de mejorar la eficiencia y precisión de estas inspecciones, reduciendo su tiempo y minimizando los errores humanos. Para ello, se ha realizado un pequeño estudio de las tecnologías actuales empleadas en la industria. A continuación, se plantea un diseño del sistema compuesto por cámara, computadora a bordo y controladora de vuelo. Seguidamente, se seleccionan los componentes comerciales necesarios para la implementación, en este caso la cámara Module 3, la computadora Raspberry Pi 4 y la controladora Pixhawk. Una vez elegidos dichos componentes, se comienza con el desarrollo del programa de control del dron mediante el lenguaje de programación Python y el uso de librerías como Dronekit. En este programa se han diseñado las distintas trayectorias que debe seguir el dron para inspeccionar toda la superficie exterior de la aeronave. Por otro lado, se ha diseñado un sistema antipérdida mediante el uso de un mallado de códigos QR situados en el suelo que deben ser detectados por la cámara situada en la parte inferior del dron. Esto permite mantener la trazabilidad del dron y la adaptabilidad a diferentes aeronaves. Finalmente, se han realizado pruebas simuladas con Dronekit-SITL para comprobar el correcto funcionamiento del programa, obteniendo resultados satisfactorios. Las conclusiones demuestran la viabilidad de realizar un software para el control de un dron autónomo capaz de realizar inspecciones visuales aeronáuticas con un bajo presupuesto. Este estudio abre la puerta a futuras investigaciones en la integración de un sistema de evitación de obstáculos para una mayor seguridad.

The use of numerical methods as a tool to solve differential equations allows for solving problems whose analytical solution is unfeasible given the current calculation techniques. Their application in the field of fluid dynamics is based on dividing a domain into a finite number of control volumes; the greater this number is, the closer the numerical solution approximates what would be the analytical solution of the problem. The Navier-Stokes equations, which express the conservation of mass, momentum, and energy of the corresponding fraction of fluid at a given moment, are applied to each control volume. This thesis provides an introduction to the numerical resolution of the Navier-Stokes equations, using a simplified formulation through certain hypothesis, with the aim of providing the author with the foundational knowledge to apply this methodology to problems in both academic and industrial fields. The results of the work demonstrate that the physical conditions of a problem, which can be quantified through the Reynolds number, or Rayleigh number in cases with natural convection, significantly affect the computational requirements to obtain its numerical solution. This numerical solution is infinitesimally close to the analytical one if grid and time step convergences are met, meaning that the number of control volumes is sufficiently high and the simulated time between one moment and the next is sufficiently small. The Reynolds number allows for establishing whether a problem corresponds to laminar (the flow is orderly and, to some extent, predictable; low value) or turbulent (the flow behaves chaotically; high value) regime. The results show a conspicuous trend towards requiring more control volumes and smaller time steps as this physical parameter increases. For this reason, the computational power can become excessive if a turbulent flow problem needs to be resolved. The most effective solution in these cases is to apply turbulence models, such as LES or RANS, instead of solving all scales of the flow, a method known as DNS.

One of the actual key challenges in the aerospace industry is reducing the weight of aircraft components while maintaining their structural integrity and functionality, resulting in extended flight range, improved payload capacity, among other benefits. Over last decades, Topology Optimization has become a innovative tool, allowing the generation of unconventional geometries that are difficult to obtain with traditional methods, respecting the performance criteria and boundary constraints. The main purpose of this thesis is to optimize the structural design of an aircraft tail stabilizer's rotational axis component using Topology Optimization techniques available on the Swan-Lab GitHub repository. In this way, the study begins with a review on present aerospace structural design, tail stabilizers and Topology Optimization techniques. Afterwards, various benchmark cases are developed, to familiarize with the SwanLab GitHub code and then the optimization characteristics of different optimizers are compared, in concrete, the Density MMA optimizer and the Level Set Nullspace one. Furthermore, an initial design of the rotational axis is performed, analysing its respective boundary conditions and load conditions according to STANAG 4671 regulations. Additionally, topological optimization is applied to redefine this design, while adjustments to the code and mesh are discussed to improve the obtained results. Lastly, the results demonstrate that the SwanLab GitHub code serves as an excellent topology optimizer with a variety of available optimizers. However, significant potential for improvement exists in areas such as time consumption, mesh inputs, and constraints related to material removal.