UPCommons (Universitat Politècnica de Catalunya)
Not a member yet
281550 research outputs found
Sort by
Design and development of control strategies for intelligent cooperative buildings in microgrids
No full textThis report contains the development and simulation of a new control strategy for intelligent cooperation between buildings in micro-grids. The project was realized in the Center of Automation and Robotics of the Spanish Superior National Research Center (CAR-CISC) as a way of using their distributed energy generation facilities. This strategy is responsible of managing the energy of each building, entailing the calculation of solar generation power, power exchange with the grid, available power for the loads and the charging/discharging of the batteries where available. The objective of the controller is to improvetheefficiencyofthewholemicro-gridbyfomentingthecooperationbetweenbuildings via power exchange. Thestrategy is developed using a systemofsystemsapproachandconsistofseveral distributed algorithms that use cooperative consensus control techniques. This means that each system of thewholesystemofsystemshasitsowncontrollerwhichmakesdecisionsbasedoninformation exchangedwiththeothercontrollers. Thecontrolarchitectureisbasedonatwo-layercontroller. Thefirstlayer is a modelpredictive controller (MPC) that calculates a reference for the system’s behavior each hour using generation and demand predictions. The second layer is a controller that operates the system in real time (in our simulations it will just be smaller intervals) by adjusting the reference given the actual values of demand and generation. This project consists on the formulation of said strategy as well as its implementation on asimulation environment using MatLAB. To utilize this environment generation and demand profiles are obtained using GridLAB-D, a simulation tool specialized in smart grids. The simulation environment is then used to evaluate the performance of this strategy under different values of the control parameters. Finally, we discuss the potential of a real-life implementation
From Alternans to Fibrillation: Simulating Cardiac Instabilities
No full textThe human heart relies on the precise propagation of electrical impulses (action potentials) to regulate its contractions. However, disruptions in this process can give rise to instabilities like rotating waves (tachycardia) or chaotic patterns (fibrillation). This study delves into the behavior of action potential dynamics using the Fenton-Karma cardiac model across different dimensions, focusing on the role of alternans, where an interaction between consecutive action potentials occurs due to an incomplete repolarization of the cell membrane.2024/20252n quadrimestr
Pla americà de Maria Nicolau durant la presentació del llibre 'Cremo!'
No full text"Tertúlia i vi" és una proposta cultural de caràcter anual organitzada per la Biblioteca del Campus Universitari de Manresa (BCUM) en col·laboració amb la DO Pla de Bages i la Llibreria Parcir. Aquesta activitat consisteix en la visita a les instal·lacions d'un dels cellers de la DO Pla de Bages, la presentació d'un llibre amb la presència del seu autor i la degustació de vins del celler amfitrió
OASST-ETC dataset: alignment signals from eye-tracking analysis of LLM responses
No full textWhile Large Language Models (LLMs) have significantly advanced natural language processing, aligning them with human preferences remains an open challenge. Although current alignment methods rely primarily on explicit feedback, eye-tracking (ET) data offers insights into real-time cognitive processing during reading. In this paper, we present OASST-ETC, a novel eye-tracking corpus capturing reading patterns from 24 participants, while evaluating LLM-generated responses from the OASST1 dataset. Our analysis reveals distinct reading patterns between preferred and non-preferred responses, which we compare with synthetic eye-tracking data. Furthermore, we examine the correlation between human reading measures and attention patterns from various transformer-based models, discovering stronger correlations in preferred responses. This work introduces a unique resource for studying human cognitive processing in LLM evaluation and suggests promising directions for incorporating eye-tracking data into alignment methods. The dataset and analysis code are publicly available.This research is supported by Horizon Europe’s European Innovation Council through the Pathfinder
program (SYMBIOTIK project, grant 101071147) and by the Industrial Doctorate Plan of the Department of Research and Universities of the Generalitat de Catalunya, under Grant AGAUR 2023
DI060Peer ReviewedPostprint (author's final draft
Jornada ‘Connecta Innovació: Universitat, Empresa i Futur’
No full textLa UPC ha acollit, el 2 de desembre, la jornada ‘Connecta Innovació: Universitat, Empresa i Futur’, destinada a reforçar els vincles entre el món acadèmic i el teixit empresarial mitjançant la creativitat i la innovació. L’acte ha comptat amb la participació de Ferran Adrià, una de les figures més influents del món gastronòmic i ambaixador de Telefónica des del 2010. Impulsada per la Càtedra Telefónica UPC-UPF, la jornada ha acollit la cloenda d’una gira que enguany també ha passat per Granada, Toledo i Las Palmas de Gran Canaria
Biological colonisation in bio-materials and composites with different bio-binder combinations
No full textThe increasing demand for sustainable and biodegradable materials in the construction industry highlights the importance of developing circular bio-based building materials that align with the principles of circular bio- economy. This study forms part of a broader investigation aimed at developing novel construction materials with suitable thermal and acoustic performance. In this context, the present work focuses on evaluating the biological colonisation susceptibility of several natural organic materials in raw form and when used with bio-binder’s, with particular attention to their fungal resistance in humid environments. The studied materials are rice husk, rice straw, Posidonia oceanica and sunflower stalk, alone and in combination with three different binders: arabic gum, xanthan gum and sodium alginate. Fungal resistance was tested exposing samples for four weeks under controlled temperature and relative humidity conditions, allowing quantification of mould growth. Results indicate that the tested unbound organic materials are highly susceptible to fungal colonisation. However, fungal proliferation could be significantly reduced by the addition of binders. Among the binders tested, sodium algi- nate showed the highest effectiveness in limiting fungal growth for all the tested materials, especially when combined with Posidonia oceanica, which presented the lowest bio-colonisation susceptibility. On the other hand, rice husk and sunflower stalk showed the highest fungal susceptibility, even with the application of binders. The results of this work contribute to a better selection of bio-materials and binders in order to enhance durability and sustainability for bio-based construction composites, contributing to the implementation of circular economy strategies in the built environment.This work was partial funded by MICIU/ AEI/10.13039/ 501100011033 through the project PID2020–117530RB-I00. Also, thanks are due to the Generalitat de Catalunya for the quality accredi- tation given to the research group GICITED (2021 SGR 01405), the Portuguese Foundation for Science and Technology support through funding UIDB/04625/2025 of the research unit CERIS and the European Union Erasmus+ Programme for the project Bio-Fibre (2022–1-DK01- KA220-HED- 00086641). The first author gratefully acknowledges the financial support of the Spanish Minister scholarship: “Subprograma Estatal de Formación del Programa Estatal para Desarrollar, Atraer y Retener Talento, en el marco del Plan Estatal de Investigación Científica.Peer ReviewedPostprint (published version
Study for the computational resolution of conservation equations of mass, momentum and energy using finite volume techniques
No full textThe following bachelor’s thesis focuses on the numerical resolution of the conservation equations of mass, momentum, and energy. The main objective of the study is to introduce the student to the field of Computational Fluid Dynamics (CFD) by solving several problems proposed by the Heat Transfer Technology Center (CTTC). Each case is simulated through a self-developed code programmed in the C++ language and the corresponding results are verified based on reference data provided by scientific articles. First, both the mathematical formulation and the numerical methods used throughout the work are introduced. In these chapters, the conservation equations are developed and the finite volume method (FVM) is described. Next, the four proposed cases are solved, presented in an orderly manner and following a consistent methodology. Each case includes the theory of the governing equations, the discretization and description of the problem, the algorithm implemented in the programmed code and, finally, the analysis and verification of the results. To conclude, a summary of the budget is included, the environmental implications of the project are described and the final conclusions of the work are presented. The complexity of the problems studied increases progressively, incorporating new concepts as each of them is explored in greater depth. It begins with a simple heat transfer case by conduction, with pure diffusion, and continues with a problem in which the convection term is analyzed by solving the general transport equation. Finally, the numerical resolution of the incompressible Navier-Stokes equations is addressed. This last part is based on the Fractional Step Method (FSM) and includes both a forced convection case, where the fluid motion is induced by boundary conditions, and a natural convection case in which the buoyancy generated by temperature gradients is responsible for the motion
On the integration of geometry agnostic variational-autoencoders into large-scale SVD based models
No full textThis manuscript introduces GAVI (Geometry Agnostic Variational-autoencoder Integration), a scalable technique for model order reduction of terabyte-level state spaces. Such methodology is agnostic to the spatial location of the data points, hence, it can handle any complex geometry regardless of the grid type in which it is represented. The dimensionality reduction of GAVI is performed on two separate steps. First, a parallel QR factorization is applied to the snapshot matrix to decompose, then a variational autoencoder learns a latent representation of the reduced R matrix. The final step can be done at laptop level regardless of the size of the original data. The methodology is used to compress the streamwise velocity component of the flow around a circular cylinder at = ∞∕ = 104 , the pressure coefficient of the flow around the Windsor body at = ∞∕ = 2.9 × 106 and the streamwise velocity of the realistic urban flow in the Zona Universitària
neighborhood located in Barcelona. The latter example is the most demanding one as the 1032 snapshots, which are represented on an unstructured grid of 335 million points, have a total size of 2.05 Tb. GAVI can compress the full dataset into 6 latent vectors that recover up to the 95.29% of the energy.The research leading to this work has been partially funded by the project TIFON with reference PLEC2023-010251/ AEI/10.13039/ 501100011033 and by the Ministerio de Ciencia e Innovación of Spain (PID2023-150408OB-C21/C22). B. Eiximeno’s work was funded by a contract from the Subprograma de Ayudas Predoctorales given by the Ministerio de Ciencia e Innovación (PRE2021-096927). Oriol Lehmkuhl has been partially supported by a Ramon y Cajal postdoctoral contract (Ref: RYC2018-025949-I). The authors acknowledge the support of Departament de Recerca i Universitats de la Generalitat de Catalunya to the Research Group Large-scale Computational Fluid Dynamics (Code: 2021 SGR 00902) and the Turbulence and Aerodynamics Research Group (Code: 2021 SGR 01051). Nathan Kutz acknowledges the NSF AI Institute in Dynamics System.Peer ReviewedPostprint (published version
Estudio y predicción de averías en tuberías de la red de distribución de agua potable
No full textL’envelliment progressiu de la xarxa de distribució d’aigua ha incrementat notablement la freqüència d’avaries en els últims anys, fet que subratlla la necessitat de desenvolupar models predictius capaços d’anticipar aquestes avaries i optimitzar les tasques de manteniment.
En aquest treball, s’avalua el rendiment de diversos algoritmes d’aprenentatge automàtic i intel·ligència artificial per a la detecció primerenca d’avaries, incorporant un enfocament col·laboratiu d’intercanvi de dades entre operadores. D’altra banda, mitjançant una anàlisi d’importància basada en valors SHAP (SHapley Additive exPlanations), es desglossa la contribució de cada variable al model, identificant la longitud de la canonada i el nombre acumulat d'avaries com a factors determinants. A més, s’investiga l’envelliment del polietilè (PE) en funció de l’antiguitat dels tubs, aplicant metodologies d’assaig destructiu i no destructiu per caracteritzar el mòdul elàstic del material. Els resultats mostren que una intervenció preventiva, reparant el 5% de les canonades, pot evitar entre el 40% – 55% de les avaries, i que la rigidesa del PE augmenta entre un 12,9% – 20,6% en tubs de 23–24 anys i un 14,4% – 71,5% en tubs de 41 anys. Aquests descobriments demostren l’eficàcia de combinar enfocaments predictius junt amb propietats mecàniques per prolongar la vida útil de la infraestructura, reduir costos operatius i augmentar el rendiment hidràulic de la xarxa.El envejecimiento progresivo de la red de distribución de agua ha incrementado notablemente la frecuencia de averías en los últimos años, lo que subraya la necesidad de desarrollar modelos predictivos capaces de anticipar estos fallos y optimizar las labores de mantenimiento.
En este trabajo, se evalúa el rendimiento de diversos algoritmos de aprendizaje automático e inteligencia artificial para la detección temprana de averías, incorporando un enfoque colaborativo de intercambio de datos entre operadoras. Por otro lado, mediante un análisis de importancia basado en valores SHAP (SHapley Additive exPlanations), se desglosa la contribución de cada variable al modelo, identificando la longitud de la tubería y el número acumulado de fallos como factores decisivos. Además, se investiga el envejecimiento del polietileno (PE) en función de la antigüedad de los tubos, aplicando metodologías de ensayo destructivo y no destructivo para caracterizar el módulo elástico del material. Los resultados muestran que una intervención preventiva, reparando el 5% de las tuberías, puede evitar entre el 40% – 55% de las averías, y que la rigidez del PE aumenta entre un 12,9% – 20,6% en tubos de 23–24 años y un 14,4% – 71,5% en tubos de 41 años. Estos hallazgos demuestran la eficacia de combinar enfoques predictivos junto con propiedades mecánicas para prolongar la vida útil de la infraestructura, reducir costes operativos y aumentar el rendimiento hidráulico de la red.The progressive ageing of the water distribution network has significantly increased the frequency of failures in recent years, underscoring the need to develop predictive models capable of anticipating these failures and optimizing maintenance tasks.
In this work, we evaluate the performance of various machine learning and artificial intelligence algorithms for early failure detection, incorporating a collaborative data‑sharing approach among operators. Meanwhile, through a SHAP (SHapley Additive exPlanations) based importance analysis, we break down each variable’s contribution to the model, identifying pipe length and cumulative failure count as decisive factors. Furthermore, we investigate polyethylene (PE) ageing, applying destructive and non‑destructive testing methodologies to characterize the material’s elastic modulus. Results show that a preventive intervention, repairing 5% of pipes, can prevent between 40% - 55% of failures, and that PE stiffness increases by 12.9% – 20.6% in 23–24‑year‑old pipes and 14.4% – 71.5% in 41‑year‑old pipes. These findings demonstrate the effectiveness of combining predictive approaches and mechanical analysis to extend infrastructure lifespan, reduce operating costs, and enhance the hydraulic performance of the network