DOCTORAL DEGREE IN AEROSPACE SCIENCE AND TECHNOLOGY

• SAMALA, BHANU PRAKASH REDDY: Computational Study of Active Flow Control in Aerodynamic Bodies: from DNS to Machine-Learning Surrogates
  
  Author: SAMALA, BHANU PRAKASH REDDY
  Programme: DOCTORAL DEGREE IN AEROSPACE SCIENCE AND TECHNOLOGY
  Department: Department of Physics (FIS)
  Mode: Normal
  Deposit date: 14/10/2026
  Reading date: 27/11/2026
  Reading time: 10:00
  Reading place: Aula 212, 2.ª planta, edificio B4, Campus Nord – UPC.
  Thesis director: MELLIBOVSKY ELSTEIN, FERNANDO PABLO | BERGADÀ GRANYÓ, JOSEP MARIA
  Thesis abstract: This thesis investigates active flow control (AFC) over aerodynamic bodies using high-fidelity direct numerical simulations (DNS) and machine-learning (ML) surrogate models. DNS provides the most accurate description of the flow physics but is feasible only at moderate Reynolds numbers, and the multi-dimensional AFC parameter space renders exhaustive parametric studies computationally prohibitive. The objective of the thesis is to advance the understanding of AFC physics across two canonical aerodynamic configurations, and to develop and evaluate ML methodologies that reduce the computational cost of high-fidelity aerodynamic analysis in industrially relevant settings.First, DNS of the flow around an Ahmed body at Re_L = 6 x 10^4 was performed with steady suction (SS) and steady blowing (SB) through a slit at the upper surface near the slant leading edge. SS reduces the drag coefficient by 8.4%, whereas SB increases it by 17.5%. The reduction under SS is attributed to flow attachment over the slant, a strengthened C-pillar vortex and pressure recovery, while SB leads to premature separation.Second, DNS of periodic zero-net-mass-flux (ZNMF) synthetic jet actuation was performed on a three-dimensional NACA 2412 airfoil at Re = 5000 and angle of attack alpha = 8 degrees, at half (sub-harmonic, SH), the same (natural harmonic, NH) and double (second harmonic, 2H) the natural vortex shedding frequency. All cases improve the lift-to-drag ratio, by 75% (NH), 47% (SH) and 18% (2H). The superior performance of NH actuation is attributed to resonant lock-in between the forcing and the dominant shedding frequency of the baseline flow, which compresses the laminar separation bubble. SH and NH actuation suppress trailing-edge spanwise vortex breakdown and reorganise the flow into quasi-two-dimensional coherent structures.Third, a physics-informed machine learning (PIML) framework was developed to correct the Reynolds stress tensor predicted by Reynolds-Averaged Navier-Stokes (RANS) solvers, using DNS-derived data and Galilean-invariant input features. It was validated on two-dimensional backward-facing step and airfoil configurations at Re ~ 3 x 10^4. Including training data that span the flow regime of the test case is a prerequisite for reliable prediction with the Random Forest algorithm. In its current form, the framework requires flow-physics similarity between training and test domains. The enhanced RANS solution is obtained in approximately one to one and a half hours, against 72 to 96 hours for DNS on four cores.Fourth, a graph-neural-network (GNN) surrogate was implemented to predict drag coefficients and surface pressure and wall shear stress distributions on industrial electric SUV geometries, using graphs derived from decimated CFD surface meshes. It was trained on RANS CFD datasets from two design campaigns comprising 136 simulation variants. Single-vehicle models reach test R^2 values of 0.625 and 0.629, and the combined model 0.932. A surface field prediction takes less than one minute, against six to eight hours on a 300-core CPU cluster for the corresponding RANS simulation.Together, the studies advance the understanding of AFC physics from first principles and provide efficient data-driven tools for industrial design.
  

DOCTORAL DEGREE IN AGRI-FOOD TECHNOLOGY AND BIOTECHNOLOGY

• BISCAMPS DALMAU, JORDI: Generació de mapes de vigor mitjançant sensors remots i proximals per a l’aplicació a dosis variable de productes fitosanitaris en vinya utilitzant tecnologia de modulació d’ample de pols (PWM).
  
  Author: BISCAMPS DALMAU, JORDI
  Programme: DOCTORAL DEGREE IN AGRI-FOOD TECHNOLOGY AND BIOTECHNOLOGY
  Department: Department of Agri-Food Engineering and Biotechnology (DEAB)
  Mode: Normal
  Deposit date: 11/09/2026
  Reading date: 03/12/2026
  Reading time: 10:00
  Reading place: Sala de Graus de l'EEABB, Campus del Baix Llobregat de la UPC, C. Esteve Terradas 8, Ed. D4, Castelldefels
  Thesis director: GIL MOYA, EMILIO | GARCIA RUIZ, FRANCISCO JOSE
  Thesis abstract: This doctoral thesis contributes to the development and validation of variable-rate technologies for the application of plant protection products in vineyards through flow-rate control systems based on pulse-width modulation (PWM) and vegetation characterization methods using remote and proximal sensors. First, a prototype based on PWM valves was developed and validated. The system used prescription maps generated from multispectral imagery acquired with an unmanned aerial vehicle to identify vigor zones and automatically adjust the flow rate. Trials were conducted in both trellised and goblet-trained vineyards at two phenological stages. The valves showed an almost linear response over a duty-cycle range of 20-80%. Under field conditions, the system reduced the spray volume applied by up to 23.8%, with an average reduction of 18.6%, without significant differences in normalized deposition compared with the conventional treatment. Second, three application strategies were compared: a conventional constant-volume system, a pressure-controlled variable-rate system, and the PWM-based system. The results showed that working pressure and the resulting droplet spectrum had a greater influence on spray drift than the total volume applied. The conventional system generated ground losses close to 40% under low-vigor conditions, whereas the PWM system reduced both ground losses and airborne drift under all the conditions studied. Finally, a device for high-throughput vegetation characterization was designed using ultrasonic sensors, GNSS, and an onboard control unit. The system was validated under laboratory and field conditions and compared with three-dimensional models generated using UAV photogrammetry. Both methods provided comparable estimates of the main structural parameters of the vineyard, with mean absolute percentage errors below 10%. Although photogrammetry offered higher spatial resolution, ultrasonic sensors proved to be a robust and operational alternative under field conditions.
  

DOCTORAL DEGREE IN ARCHITECTURAL, BUILDING CONSTRUCTION AND URBANISM TECHNOLOGY

• REYNA RUIZ, CATALINA: Producción de bloques de tierra comprimida (BTC) para climas cálidos y húmedos con maquinaria adobera, y el establecimiento del agente estabilizador óptimo para el cumplimiento de los códigos locales.
  
  Author: REYNA RUIZ, CATALINA
  Programme: DOCTORAL DEGREE IN ARCHITECTURAL, BUILDING CONSTRUCTION AND URBANISM TECHNOLOGY
  Department: Department of Architectural Technology (TA)
  Mode: Article-based thesis
  Deposit date: 21/09/2026
  Reading date: 12/11/2026
  Reading time: 16:30
  Reading place: ETSAB (Esc.Téc.Sup.Arquit.Bcn)-Pl. Baja-Sala GradosAv. Diagonal, 649Enlace a videoconferencia: https://meet.google.com/nyr-vdqm-jboConnexión 16.00h
  Thesis director: GOMEZ SOBERON, JOSE MANUEL VICENTE | ROJAS VALENCIA, MARIA NEFTALI
  Thesis abstract: ABSTRACTAs world population grows so does the already existing housing deficit and it becomes evident that the construction industry has a responsibility to mitigate its environmental impacts that come from the direct extraction of raw materials, their transformation and the transportation. On the other hand, the implementation of climate control systems increases the energy demand.Simultaneously there is a global increase of the production of different waste materials that if not treated pose a great challenge to maintain ecosystems healthy. In the case of the United States, the annual volume of the construction and demolition residues as well as the glass that are not recycled is equivalent to 20,000 and 546 bases of the Statue of Liberty respectively.Return the gaze to the earth, is not a beautiful metaphorical phrase but a proposal for the construction industry to opt for materials with a reduced carbon footprint that can be sourced locally to reduce it transportation impact and that offer a greater insulation capacity. Those are the main characteristics of the Compressed Earth Blocks (CEB). If these include closed loop materials in their formulation; a greater sustainability is achieved.This research work studied the incorporation of local soils and aggregates as well as two closed loop materials in the production of Compressed Stabilized Earth Blocks (CSEB). One group with lime and another with cement in three different concentrations by weight (5, 10 and 15%). They were manufactured with a semi-automatic machine in a locality that had average temperature of 32 °C and 91% relative humidity. Four matrices were design per type of stabilizer, the first one was considered the one of reference and included the soils and the natural aggregates, for the second one, natural sand was replaced with recycled glass sand, for the third one the natural aggregates were replaced with the crushed concrete potion of construction and demolition wastes and the fourth one incorporated both closed loop materials (24 formulations). In the case of the CSEB with lime, 92% of the formulas exceeded the simple compression test (SCT) and capillary water absorption (AbsCoeff10) values established by the normativity. The highest SCT obtained was 4.3 MPa. For the CSEB with cement, a broader set of characteristics were studied; such as apparent porosity (AP) – immersion test, modulus of elasticity (MoE) and thermal conductivity (TC) in the second phase of the research and additionally, for the third one; thermogravimetric analysis (TGA), scanning electron microscopy (SEM), and spectrometry of dispersed X-ray energy (EDS). With an analysis of optical images (OIA) a method was presented to correlate the formulas with some of the properties, including predicting capabilities with mathematical equations. Analyzing SEM images ITZ width values and porosity values were obtained. The greatest SCT value registered was 7.3 MPa and the lowest values of AbsCoeff10 and CT were 8 g/(cm² × min0.5) y 0.684 W/m·K. The formulas with 15% of cement showed the biggest mass loss, associated to the presence of interlayer water and the hydration products as well as the greater porosity and ITZ width reduction with the highest concentrations of silicon and calcium.
  

DOCTORAL DEGREE IN ARCHITECTURAL, CIVIL AND URBAN HERITAGE AND REFURBISHMENT OF EXISTING BUILDINGS

• GARCIA SALANOVA, ANTONIO: Arquitectura amb propòsit per a una gestió sostenible dels equipaments públics
  
  Author: GARCIA SALANOVA, ANTONIO
  Programme: DOCTORAL DEGREE IN ARCHITECTURAL, CIVIL AND URBAN HERITAGE AND REFURBISHMENT OF EXISTING BUILDINGS
  Department: Departamento de Representación Arquitectónica (RA)
  Mode: Normal
  Deposit date: 09/09/2026
  Reading date: 18/11/2026
  Reading time: 10:30
  Reading place: ETSAB (Escuela Técnica Superior de Arquitectura de Barcelona) - Planta -1 - Aula C-B11Av. Diagonal, 649-651 - 08028 - Barcelona
  Thesis director: REDONDO DOMINGUEZ, ERNEST
  Thesis abstract: PURPOSEFUL ARCHITECTURE FOR THE SUSTAINABLE MANAGEMENT OF PUBLIC FACILITIESThis research analyses, formulates and reasonably contrasts a comprehensivea model of sustainable management of public facilities throughout their life cycle, from planning and design to operation and dismantling, and incorporates data-driven governance aimed at facilitating collaborative decision-making between the different agents involved. It is based on the hypothesis that the adoption of a comprehensive sustainable management model, based on collaborative digital methodologies —especially Building Information Modeling (BIM)—, sustainability criteria and life cycle analysis (LCA), makes it possible to overcome the limitations of traditional models of management of public facilities, characterized by the fragmentation of processes, the lack of traceability of information and the low optimization of the use phase. The work adopts a mixed methodological approach that combines systematic review of the literature, comparative analysis between management models, interviews with experts from the public and private sectors, quantitative validation and the development of an original methodological instrument: a methodological framework of 100 sustainable technical criteria applicable to the entire life cycle, which operationalizes the proposed management model. Unlike predominantly theoretical approaches, the research is carried out in a real context of public management of educational facilities and incorporates recent data and professional experiences (2024–2025), especially in relation to the implementation of BIM in the operation phase.The research also develops a practical case, a hypothetical Audiovisual Vocational Training Institute of Catalonia in the Tres Xemeneies del Besòs site, which acts as a proof of concept of the proposed comprehensive management model. This case makes it possible to contrast the coherence and feasibility of the model, as well as its architectural translation in relation to the pedagogical requirements of vocational education.The results positively show the three hypotheses formulated. Firstly, the comprehensive management model based on BIM–LCA improves the efficiency, quality and sustainability of processes. Secondly, its application determines a new flexible, modular architectural typology aimed at efficient maintenance. Thirdly, the integral model is adapted in a coherent way to the functional and pedagogical requirements of Vocational Training, reinforcing the connection between architecture, management, educational innovation and participatory governance processes. Of particular relevance is the contribution in the field of digital operation management, still underdeveloped in current institutional practice,Finally, the research proposes future lines aimed at data-driven exploitation, the integration of intelligent systems and the development of new professional profiles linked to life cycle management.
  

DOCTORAL DEGREE IN AUTOMATIC CONTROL, ROBOTICS AND VISION

• CRUZ RANGEL, DAVID ANTONIO: Electrical-Current-based Physical Feature Extraction and Online Condition Assessment for Rotodynamic Systems Monitoring
  
  Author: CRUZ RANGEL, DAVID ANTONIO
  Programme: DOCTORAL DEGREE IN AUTOMATIC CONTROL, ROBOTICS AND VISION
  Department: Department of Automatic Control (ESAII)
  Mode: Normal
  Deposit date: 17/09/2026
  Reading date: 06/11/2026
  Reading time: 11:00
  Reading place: Aula 28.8, Escola Tècnica Superior d'Enginyeria Industrial de Barcelona (ETSEIB) Campus Diagonal Sud, Av. Diagonal, 647 08028 Barcelona
  Thesis director: OCAMPO MARTINEZ, CARLOS AUGUSTO | DÍAZ ROZO, JAVIER
  Thesis abstract: This dissertation investigates the use of electrical current measurements as a non-intrusive observation tool for industrial systems. The work is motivated by the need for scalable, cost-effective, and online monitoring solutions able to extract meaningful information from industrial assets without extensive dedicated instrumentation. The thesis focuses on rotodynamic systems and nonlinear electromechanical processes, where mechanical, hydraulic, electrical, and process-related phenomena are indirectly embedded in current signals.Conventional condition monitoring commonly relies on dedicated sensors such as accelerometers, pressure transducers, flow meters, acoustic sensors, or temperature probes. Although these sensors provide valuable information, their deployment may increase cost, complexity, maintenance requirements, and intrusiveness, particularly when many assets must be monitored continuously. In contrast, electrical current can be acquired non-intrusively using clamp-on sensors and is widely available in electrically driven systems, making it a promising information source for industrial monitoring.The central hypothesis of this thesis is that current signals contain information beyond energy consumption. Through electromechanical coupling, variations in mechanical load, hydraulic disturbances, operating regimes, degradation processes, and process dynamics can modify the electrical response of the system. Therefore, with suitable signal-processing, feature-extraction, and condition-assessment methods, current measurements can be transformed into a practical observation modality for online monitoring.The thesis is organized into four parts. Part I introduces the motivation, research questions, and state of the art in online Non-Intrusive Load Monitoring (NILM) and current-based industrial monitoring, emphasizing limitations related to physical interpretability, non-stationary operation, computational cost, scalability, and edge deployment. Part II develops the core methodologies for physical feature extraction, characterization, and identification from current measurements. First, a PSD-based current signature analysis method is proposed for variable-speed rotodynamic systems, enabling the extraction of mechanically meaningful degradation indicators under non-stationary conditions. The framework is then extended to hydraulic phenomena in centrifugal pumps by combining positive-sequence magnitude analysis and rotational-frequency-normalized PSD representations for cavitation detection. Finally, a non-intrusive framework is developed for coupled electromechanical systems operating under aggregated electrical measurements, enabling load identification, activation tracking, and health-state assessment.Part III integrates the proposed methodologies into online monitoring frameworks for industrial deployment. An online current-signature-based condition assessment approach is developed for HVAC water pumps, demonstrating the feasibility of cavitation monitoring under real operating conditions. The final framework is deployed in an industrial hydroelectric turbine, where electrical, mechanical, and hydraulic phenomena are monitored using only three-phase current measurements. This deployment demonstrates the practical applicability of the proposed methodology within an edge-computing architecture and under realistic constraints.Overall, this dissertation advances a unified, scalable, and physically interpretable current-based monitoring framework for industrial systems. The developed methodologies show that electrical current measurements can go beyond energy monitoring or load identification, serving as a rich source of information about mechanical degradation, hydraulic anomalies, process dynamics, and condition assessment. These findings support current signals as a versatile, non-intrusive diagnostic tool, especially where traditional sensors are costly, impractical, or difficult to maintain.
  
• GEBELLÍ GUINJOAN, FERRAN: User-centred Explainable Human-Robot Interaction
  
  Author: GEBELLÍ GUINJOAN, FERRAN
  Programme: DOCTORAL DEGREE IN AUTOMATIC CONTROL, ROBOTICS AND VISION
  Department: Institute of Robotics and Industrial Informatics (IRI)
  Mode: Normal
  Deposit date: 15/09/2026
  Reading date: 17/11/2026
  Reading time: 10:00
  Reading place: Sala de juntes de la Facultat de Matemàtiques i Estadística (FME), Campus Diagonal Sud, Carrer de Pau Gargallo, 14, 08028 Barcelona
  Thesis director: ROS ESPINOZA, RAQUEL | LEMAIGNAN, SÉVERIN
  Thesis abstract: As robots are increasingly deployed in everyday human environments, the ability to explain their actions and decisions to users becomes essential for building trust, fostering collaboration, and enabling effective interaction. However, existing approaches to explainability in Human-Robot Interaction (HRI) often treat users as a homogeneous group, neglecting the diversity in knowledge, cognitive styles, and informational needs that characterise real-world deployments. This thesis investigates user-centred explainability in HRI through a series of longitudinal, in-the-wild studies that span the full lifecycle of explanation design, evaluation, generation, and personalisation. First, a participatory design methodology has been developed and implemented in a real healthcare use case, demonstrating that user expectations and requirements for explainability emerge gradually during in-situ deployment, and that involving users in the early design stages is crucial to incorporate a user-centred perspective on explainability. Secondly, we have explored the evaluation of explainability, focusing on one of the main desired outcomes of explainability for users: improving their understanding of the robots they interact with. We present a systematic literature review which identifies significant variability in how understandability is conceptualised and measured in HRI, highlighting the lack of standardised tools and longitudinal, in-the-wild evidence. This literature review also poses an open question about the relationship between objective and subjective understandability measures. Because of that, we conduct a two-month field study that demonstrates that objective and subjective understandability can evolve independently over time: users may report an increase in perceived understanding while the actual comprehension remains poor, underscoring the need to assess both dimensions separately. Finally, this thesis addresses the need for user-centred explanation generation, making use of Large Language Models (LLMs). We first introduce a modular framework that orchestrates specialised component explainers mapped to individual robot subsystems, which produces more accurate and contextually appropriate explanations compared to monolithic approaches. Building on this, and linking with the thesis topic and previous findings, a personalisation framework is developed that adapts explanation detail to each user by tracking acquired knowledge from previously received explanations over time. A one-month office deployment validates that personalised explanations are perceived as more appropriately detailed and remain concise, although their benefits depend on the criticality of what is being explained and diminish over extended periods, potentially due to predictability costs and anchoring effects. Taken together, these contributions advance the understanding of explainability as a dynamic, user-dependent property and provide concrete methods and empirical evidence to guide the design and development of explainable human-robot interaction in real-world settings.
  
• PEÑA CAMPOS, JOHAN SEBASTIAN: Interpretable Decision Support Systems for Early Cancer Diagnosis and Personalized Treatment
  
  Author: PEÑA CAMPOS, JOHAN SEBASTIAN
  Programme: DOCTORAL DEGREE IN AUTOMATIC CONTROL, ROBOTICS AND VISION
  Department: Department of Automatic Control (ESAII)
  Mode: Change of supervisor
  Deposit date: 22/09/2026
  Reading date: 04/12/2026
  Reading time: 11:00
  Reading place: Aula 28.8 del ETSEIB, Escola Tècnica Superior d'Enginyeria Industrial de Barcelona (ETSEIB) Campus Diagonal Sud, Av. Diagonal, 647, 08028 Barcelona
  Thesis director: OCAMPO MARTINEZ, CARLOS AUGUSTO | PATIÑO GUEVARA, DIEGO ALEJANDRO
  Thesis abstract: Cancer remains a leading cause of death worldwide, with gastric and metastatic prostate cancer imposing substantial burdens, particularly in low- and middle-income countries. This thesis develops interpretable Decision Support Systems (DSS) for two phases of cancer care: early diagnosis and personalized treatment optimization.The first stage targets early diagnosis of gastric cancer using routinely collected data from the Colombian health system (SGSSS). A deep learning architecture combines bidirectional Long Short-Term Memory (LSTM) networks and multi-head attention to process sequences of diagnostic codes and procedures from over 4.9 million patient trajectories, with Spanish biomedical embeddings, intelligent undersampling and Focal Loss addressing the extreme class imbalance (5103:1). It is then reformulated as a more parsimonious 2D convolutional network (CNN-2D) with demographic features, adopted as the final model for its interpretability, parsimony (~176k vs ~4.5M parameters) and competitive performance (AUC-ROC ≈ 0.835). The Nonlinear Oblique Subspace Projections (NObSP) methodology is extended to dynamic systems to interpret the model, decomposing predictions into clinically meaningful components (reconstruction error below 5%) and separating genuine clinical signal, an enriched upper-gastrointestinal pattern and a coherent age threshold, from geographic and utilization confounds.The second stage addresses personalized treatment for metastatic castrate-resistant prostate cancer (mCRPC) by integrating Evolutionary Game Theory (EGT) and Model Predictive Control (MPC). Lotka-Volterra competition models capture frequency-dependent selection among phenotypically distinct cell populations under therapy, and MPC optimizes adaptive schedules balancing tumor burden, resistance, toxicity and parsimony. For single-agent Abiraterone, it extends the median Time to Progression (TTP) from 275 days under continuous Maximum Tolerated Dose (MTD) to 420 days, roughly a 50% improvement over continuous MTD, while using about one fifth of its cumulative drug dose, and it outperforms statically optimized schedules (344 and 340 days). Dual-agent formulations coordinating Abiraterone and Docetaxel delay resistant-population dominance beyond 8,000 days, about 50% longer than the best adaptive-therapy baseline of West et al., while a blocking strategy cuts computation by over 98% versus recomputing at every step.The framework is further extended by introducing replicator dynamics, the canonical equation of EGT, as an alternative prediction model, deliberately creating a structural mismatch with the Lotka-Volterra plant while keeping the controller, objective and decision criteria unchanged. The replicator-based controller yields smoother single-agent dosing with equivalent clinical outcomes, and a minimal intervention, an 81% reduction in cumulative exposure relative to the matched Lotka-Volterra controller, suffices for sustained tumor control over the full simulation horizon in the dual-agent case, with 94% less computation than that controller. This gain follows from the modeling choice, not a different control law.Both stages treat interpretability as a core design requirement, operate within resource-constrained settings and establish validation frameworks that open pathways toward clinical evaluation, showing how mathematical and computational methods can be applied systematically across cancer care to improve patient outcomes.
  

DOCTORAL DEGREE IN BIOMEDICAL ENGINEERING

• CASTELLS DELGADO, ANTONI: Rhythmic Emotional Stimulation and Heart Rate Variability: Exploring the Limits of Visual and Acoustic Modulation of Stress
  
  Author: CASTELLS DELGADO, ANTONI
  Programme: DOCTORAL DEGREE IN BIOMEDICAL ENGINEERING
  Department: Department of Electronic Engineering (EEL)
  Mode: Normal
  Deposit date: 30/09/2026
  Reading date: 09/12/2026
  Reading time: 15:00
  Reading place: Aula de Graus de l'Escola Tècnica Superior d'Enginyeria de Telecomunicació de Barcelona (ETSETB), C4,C4-002, C/Jordi Girona, 1-3, 08034 Barcelona
  Thesis director: GARCIA GONZALEZ, MIQUEL ANGEL | FERNANDEZ CHIMENO, MIREYA
  Thesis abstract: This study investigates the effects of Rhythmic Visual Emotional Stimulation (RVES) and Rhythmic Acoustic Emotional Stimulation (RAES) on Heart Rate Variability (HRV), specifically analyzing SDNN, RMSSD, and IBI as key physiological indicators of autonomic function. HRV is a well-established biomarker of stress and autonomic regulation, and previous research has demonstrated that rhythmic external stimuli presented at the resonant frequency of the arterial baroreflex (0.1 Hz) can enhance HRV. RVES, originally developed by Vaschillo et al. (2008), utilizes emotionally charged visual stimuli to elicit high-amplitude HRV oscillations. This study extends prior research by exploring whether RAES, using emotionally matched acoustic stimuli, can induce comparable effects.Thirty healthy participants (15 males, 15 females) underwent both RVES and RAES protocols, with HRV measured before, during, and after each intervention. The results revealed distinct patterns in autonomic responses to the two modalities. RVES significantly increased SDNN, indicating its effectiveness in enhancing overall HRV variability, with females exhibiting stronger and more sustained HRV responses than males. RMSSD showed an initial increase during RVES but declined post-intervention, particularly in males, suggesting a short-term parasympathetic withdrawal. RAES, in contrast, had a weaker effect on HRV, producing only a modest increase in RMSSD during the intervention with no significant changes in SDNN. Additionally, BMI was found to be a moderating factor, with a significant three-way interaction between stimulation, sex, and BMI (p = 0.023) for IBI in RAES, highlighting individual variability in autonomic responses.These findings suggest that RVES is a more effective tool for autonomic modulation compared to RAES, supporting previous studies on visual rhythmic entrainment of the baroreflex. The weaker effects of RAES indicate that auditory stimuli alone may not be as effective as visual stimuli in eliciting sustained HRV changes. The significant influence of sex and BMI further emphasises the importance of individual differences in autonomic regulation. Future research should explore multimodal integration (visual-auditory rhythmic stimulation), personalised stimulus design, and repeated exposure effects to optimise HRV-based interventions for stress management and autonomic health.
  
• CLARAMUNT MOLET, MIREIA: Development of a novel clinical gait assessment tool for monitoring and stratifying neuromuscular diseases
  
  Author: CLARAMUNT MOLET, MIREIA
  Programme: DOCTORAL DEGREE IN BIOMEDICAL ENGINEERING
  Department: Department of Mechanical Engineering (EM)
  Mode: Normal
  Deposit date: 24/07/2026
  Reading date: 17/11/2026
  Reading time: 10:30
  Reading place: Aula 28.8. Planta 1, Escola Tècnica Superior d'Enginyeria Industrial de Barcelona (ETSEIB), Av. Diagonal, 647, 08028 Barcelona
  Thesis director: FONT LLAGUNES, JOSEP MARIA | IDELSOHN ZIELONKA, SEBASTIAN
  Thesis abstract: Neuromuscular diseases are chronic and degenerative conditions of the skeletal musculature that progressively impair gait, one of the primary determinants of functional independence. Their monitoring, both in clinical practice and clinical trials, relies on clinical scales and functional tests that are subject to observational bias, ceiling effects, and insufficient sensitivity to detect subtle changes. At a time when novel disease-modifying therapies are emerging, the availability of an objective, portable tool integrable into routine clinical practice for biomechanical mobility assessment represents an unmet clinical need.To address this need, the present thesis develops the biomechanical algorithms to derive spatiotemporal parameters, sagittal joint kinematics, muscle activation, plantar pressure distribution, and heart rate from raw wearable sensor data. Data capture relies on the Ephion Mobility system, which integrates and synchronises inertial sensors, surface electromyography, plantar pressure insoles, and a cardiac sensor within a single mobile application. Using this system, data were collected from 176 volunteers, 90 patients and 86 controls, across two countries and standard hospital settings.The first study, conducted in collaboration with Hospital Sant Joan de Déu, serves as the proof of concept. It validates a multidimensional gait score for Duchenne muscular dystrophy, derived from multisensor data through unsupervised learning, which demonstrates good agreement with current clinical metrics and promising capacity for disease monitoring.With the validity of the approach established, the second study, carried out at the Copenhagen Neuromuscular Center, explores its applicability in two conditions with markedly different distributions of muscle weakness: Becker muscular dystrophy and myotonic dystrophy type 1. Biomechanical gait assessment is shown to discriminate disease-specific patterns and severity gradients that functional tests do not capture. Finally, the third study, conducted in collaboration with the Spanish Pompe Patient Association, addresses the most demanding question: whether the tool can detect longitudinal progression in a slowly evolving disease. The Pompe-MDPI achieves statistical significance over one year of follow-up in a context where current metrics fail to detect change over the same interval.Taken together, the results provide preliminary evidence that portable multidimensional biomechanical assessment is clinically feasible, sensitive to functional severity, and capable of detecting longitudinal disease progression where conventional measures fall short. The derived scores represent promising candidates as digital biomarkers for natural history studies and clinical trials of disease-modifying therapies.
  
• GARCÍA HERNANDO, GABRIEL: Development of an advisory system for the administration of anaesthetics and analgesics
  
  Author: GARCÍA HERNANDO, GABRIEL
  Programme: DOCTORAL DEGREE IN BIOMEDICAL ENGINEERING
  Department: Department of Automatic Control (ESAII)
  Mode: Normal
  Deposit date: 18/09/2026
  Reading date: 27/11/2026
  Reading time: 11:00
  Reading place: Aula 28.8, Escola Tècnica Superior d'Enginyeria Industrial de Barcelona (ETSEIB), Campus Diagonal Sud, Av. Diagonal, 647 08028 Barcelona
  Thesis director: VALLVERDU FERRER, MARIA MONTSERRAT | GONZÁLEZ PIJUAN, CARMEN
  Thesis abstract: Current technology provides valuable indices of hypnotic state, and although adverse anaesthesia effects have declined, important challenges remain. In particular, distinguishing between pharmacologically distinct anaesthetic agents and reliably predicting and detecting intraoperative nociception from physiological signals would support more personalized perioperative care.Electroencephalography (EEG) and heart rate variability (HRV), derived from RR intervals, provide complementary information on brain and autonomic responses to anaesthetic drugs and noxious stimuli. However, the optimal parameterisation of nonlinear descriptors, such as time delay, symbolic encoding, frequency band, and analytical signal representation, has not been systematically explored for drug discrimination and nociception assessment.This thesis pursues two objectives: first, discrimination between three pharmacological states, no drug (ND), propofol (P), and remifentanil (R), using EEG-based symbolic dynamics and Poincaré plot analysis; and second, prediction and detection of intraoperative nociception, defined as movement response to a noxious stimulus, using Poincaré analysis of EEG and RR signals.A total of 599 patients under general anaesthesia were analysed. For drug discrimination, 84 patients met stability criteria (28 ND, 28 P, 28 R). For nociception analysis, 163 patients yielded 329 stimulus instances (159 movers, 170 non-movers). At the patient level, 55 were consistent movers, 72 consistent non-movers, and 36 showed mixed responses.EEG signals were decomposed into seven frequency bands. For each band, three Hilbert-based representations were derived: amplitude, envelope, and phase. The first derivative was also computed for symbolic dynamics. Poincaré descriptors and symbolic features were extracted across all bands and representations, forming a multidimensional feature space applied consistently to both analyses.Symbolic dynamics parameters (number of symbols, word length, quantification factor, sampling rate) were systematically optimized. For Poincaré analysis, time delay was exhaustively evaluated across traditional, segmented (SPPA), and asymmetry-based descriptors, applied to all EEG representations and RR windows (30–60 s). Nonlinearity was validated using surrogate data. Linear EEG and HRV features were also assessed. Quadratic discriminant analysis with leave-one-out cross-validation was used throughout. In all cases, combining features across modalities and representations outperformed single-feature approaches.For drug discrimination, symbolic dynamics achieved the highest performance (AUC = 0.952 ND vs P; 0.919 P vs R; 0.837 ND vs R). Poincaré descriptors provided complementary results: traditional descriptors performed best for ND vs P (AUC = 0.836, F1 = 84.2%) and ND vs R (AUC = 0.835, F1 = 84.7%), while asymmetry-based descriptors were optimal for P vs R (AUC = 0.893, F1 = 90.0%).For nociception, prediction was best achieved using SPPA descriptors of RR intervals over a 30 s pre-stimulus window (AUC = 0.691, F1 = 68.3%), while detection was best with traditional Poincaré descriptors of EEG (AUC = 0.700, F1 = 72.9%).Overall, nonlinear analysis of EEG and RR signals, combined with systematic parameter optimisation, enables robust drug discrimination and meaningful nociception assessment from a single frontal EEG channel. Although nociception performance was moderate, reflecting the complexity of predicting behavioural responses under anaesthesia, the consistent benefit of multimodal and multi-representation features supports their use in next-generation personalized anaesthesia monitoring systems.
  
• NAVARRO PONS, MARC: Smart materials and resonant microelectromechanical systems for bioelectronic microdevices and electroceuticals
  
  Author: NAVARRO PONS, MARC
  Programme: DOCTORAL DEGREE IN BIOMEDICAL ENGINEERING
  Department: Department of Electronic Engineering (EEL)
  Mode: Normal
  Deposit date: 17/09/2026
  Reading date: 06/11/2026
  Reading time: 11:00
  Reading place: Sala d'Actes Pepe Millán de l'Institut de Microelectrònica de Barcelona (IMB-CNM-CSIC), C. dels Til·lers, 08193, Bellaterra (Campus UAB)
  Thesis director: MURILLO RODRÍGUEZ, GONZALO
  Thesis abstract: This thesis presents the design, fabrication, characterization, and biological assessment of two complementary piezoelectric platforms for wireless electrical cell stimulation under ultrasound actuation: aluminum nitride (AlN)-based MEMS resonators (Part I) and electrospun silk fibroin nanofiber scaffolds and microparticles (Part II). The central question is whether ultrasound-actuated piezoelectric microdevices can generate extracellular electric fields of sufficient magnitude and spatial confinement to elicit measurable cellular responses at acoustic intensities within commonly accepted diagnostic-ultrasound exposure ranges.The AlN platform is based on circular clamped membrane resonators (~90 µm diameter) fabricated on SOI substrates using Pt electrodes. A total of 24 resonators are integrated in a 3x8 array on a 1x1 cm chip, each individually routed to its own stimulation electrode pair. The devices resonate at ~3.27 MHz in air (Q = 199) and at ~1.62 MHz immersed with one side under water (Q = 13). The AlN thin film exhibits d33,f = 1.4 pm/V, e31,f = -0.34 C/m2, and εr = 10.64, as measured by DBLI with substrate clamping correction. Impedance spectroscopy revealed a low coupling coefficient (k2eff ≈ 0.5%) and dominant parasitic capacitance from the routing traces (Cpar ≈ 8.5 pF) requiring a custom polynomial baseline subtraction method for BVD parameter extraction. The acoustic sensitivity under ultrasound actuation at 1 MHz was 11.2 mV/MPa, linear across 0.1 to 2.0 W/cm2. A COMSOL model, calibrated and compared with the experimental data, predicted electric fields above the field reported to induce Ca2+ responses due to VGCC activation in osteoblast-like cells at 0.3 W/cm2.Biological validation with Saos-2 cells established a controllable, intensity-dependent operating window: cell viability exceeded 95% at 0.1 and 0.5 W/cm2, dropping to approximately 10% at 1.0 W/cm2. This sharp transition is consistent with device-mediated electromechanical activation rather than the acoustic intensity alone. Calcium imaging at 0.3 W/cm2 showed that 20 of 34 cells (59%) on exposed electrodes exhibited calcium transients, compared with ~12% off-electrode. To our knowledge, this is the first cell-scale, resonant AlN-based microdevice for direct wireless electrical cell stimulation without external rectification or amplification circuitry.The silk fibroin trades the spatial precision and remote activation control for full biodegradability and distributed piezoelectric excitation. Electrospinning from aqueous silk fibroin solutions at (21 wt%) produced random and aligned nanofiber mats. Methanol immersion was selected among three crystallization methods for its high β-sheet content (~37-38% by FTIR amide I deconvolution, from ~27% as-spun), preserved fiber morphology, and reproducibility. Crystallization increased both stiffness and strength (E = 179 MPa, UTS = 6.1 MPa for random mats; E = 206 MPa, UTS = 11.1 MPa for aligned mats) and the piezoelectric coefficient 3.3-fold, from d33 = 2.2 pC/N to 7.3 pC/N, confirming β-sheet domains as the structural origin of the piezoelectric response. Crystallized films were converted into 130x130 µm microparticles by silicon microcutting at DTU Health Tech, preserving the nanofiber architecture.Biological validation demonstrated cytocompatibility with HaCaT, NHDF, and Saos-2 (viability >90%), proliferation kinetics comparable to glass controls, CK10 differentiation on the scaffold (~80%, comparable to glass) and cell adhesion and viability on the microparticle format.The two platforms are complementary: the AlN resonators provide controlled, individually addressable stimulation with defined frequency, intensity, spatial precision, and CMOS compatibility; the silk fibroin platform provides a biodegradable scaffold with distributed piezoelectric excitation and a microparticle format for tissue integration, where biodegradability and scaffold function take priority over electrical precision.
  

DOCTORAL DEGREE IN CHEMICAL PROCESS ENGINEERING

• MARÍN RIVAS, ENRIQUE DAVID: Mechanochemical Preparation of Pd-Cu/CeO₂ Catalysts for Carbon Monoxide Oxidation
  
  Author: MARÍN RIVAS, ENRIQUE DAVID
  Programme: DOCTORAL DEGREE IN CHEMICAL PROCESS ENGINEERING
  Department: Department of Chemical Engineering (EQ)
  Mode: Normal
  Deposit date: 31/07/2026
  Reading date: 13/11/2026
  Reading time: 11:00
  Reading place: Sala Polivalent - Edifici IEEBEhttps://eebe.upc.edu/ca/lescola/com-arribar
  Thesis director: LLORCA PIQUE, JORDI | VENDRELL VILLAFRUELA, XAVIER
  Thesis abstract: Cerium oxide (CeO2)-based catalysts are widely studied for oxidation reactions due to their redox properties, oxygen storage capacity, and ability to stabilise metal-support interfacial sites. Among these systems, Pd/CeO2 and Cu/CeO2 catalysts are of particular interest for low-temperature CO oxidation. However, their catalytic behaviour depends on the synthesis method, metal dispersion, and defect structure. In this thesis, mechanochemical synthesis by planetary ball milling was investigated as a sustainable route for preparing CeO2-based catalysts, together with different treatments aimed at modifying their structural and catalytic properties.The first part of this work focused on Pd/CeO2 catalysts prepared by planetary ball milling under controlled atmospheres. The results demonstrated that mechanochemical synthesis promotes the formation of active Pd-Ce interfacial structures. Catalysts prepared at 250 rpm under air for 10 min exhibited the highest CO oxidation activity, outperforming those prepared by conventional impregnation. This behaviour was attributed to the formation of PdOx-Ce interfacial species and strong metal-support interactions. Electron microscopy revealed highly dispersed Pd nanoparticles and clusters on the ceria surface.The second part investigated Cu/CeO2 catalysts prepared by mechanochemical methods using different precursors, calcination temperatures, milling conditions, and reduction treatments. The results showed that mechanochemical synthesis can produce catalysts with superior performance compared with conventional methods. Catalysts prepared from copper nitrate and calcined at 300 °C showed the best activity. H2/N2 reduction further enhanced catalytic performance. Electron microscopy confirmed the absence of large CuOx aggregates and the presence of highly dispersed copper species. In addition, operando synchrotron experiments revealed dynamic Cu+/Cu2+ redox cycling under reaction conditions, contributing to catalyst stability and progressive activation.The third part addressed bimetallic Pd-Cu/CeO2 catalysts prepared by planetary ball milling. Different Cu/Pd ratios, metal loadings, and deposition sequences were evaluated. Under non-reduced conditions, Pd incorporation did not significantly improve activity compared with monometallic Cu/CeO2 catalysts. However, after H2 reduction, the bimetallic catalysts exhibited enhanced activity, highlighting the importance of pre-reduction in the formation of favourable Pd-Cu-Ce interfacial structures.Electron beam irradiation was also investigated as a post-synthesis treatment for Pd/CeO2 catalysts. Irradiation improved catalytic activity and reduced the apparent activation energy for CO oxidation. Raman and XPS analyses revealed an increase in structural defects, peroxide species, and additional PdOx-Ce interfacial sites.Finally, flash sintering was applied to ceria-based materials. Although the treatment induced significant microstructural and electrical modifications, the resulting materials exhibited lower catalytic activity than untreated CeO2, mainly due to the thermal effects associated with Joule heating.Overall, this thesis demonstrates that mechanochemical synthesis is an effective strategy for preparing highly active CeO2-based catalysts. The catalytic behaviour of Pd/CeO2, Cu/CeO2, and Pd-Cu/CeO2 systems is governed by metal-ceria interfacial structures, metal dispersion, and the redox properties of the support.
  
• MOGHADAMFAR, TANAZ: Advanced membrane modification strategies to improve the ion selectivity of nanofiltration membranes for developing circular processing of brackish water
  
  Author: MOGHADAMFAR, TANAZ
  Programme: DOCTORAL DEGREE IN CHEMICAL PROCESS ENGINEERING
  Department: Department of Chemical Engineering (EQ)
  Mode: Article-based thesis
  Deposit date: 22/09/2026
  Reading date: 06/11/2026
  Reading time: 10:30
  Reading place: EEBE - UPCSala d'actes, Edifici A
  Thesis director: REIG I AMAT, MÒNICA | DEL VALLE MENDOZA, LUIS JAVIER
  Thesis abstract: Water treatment has become a critical global challenge due to increasing water scarcity while supporting the European Union's transition toward a circular economy and the objectives of the European Green Deal. Besides water reuse, treatment technologies are increasingly expected to recover valuable resources, including critical raw materials such as Mg²⁺, from water and wastewater streams. Membrane technology plays a central role in this transition because of its energy efficiency and high selectivity toward ionic species. Among membrane-based processes, nanofiltration (NF) has emerged as a promising technology for water treatment and selective ion separation. However, conventional NF membranes still suffer from limited ion selectivity, membrane fouling, and insufficient control of ion transport, particularly for challenging separations between monovalent and divalent ions.This thesis focused on the development of advanced NF membranes through surface modification and nanocomposite fabrication strategies to improve ion selectivity, permeability, and stability. The main approach involved layer-by-layer (LBL) assembly using poly(diallyldimethylammonium chloride)/poly(styrene sulfonate) (PDADMAC/PSS) and poly(allylamine hydrochloride)/poly(styrene sulfonate) (PAH/PSS). The influence of polyelectrolyte type, concentration, and the number of bilayers (1.5–6.5) was systematically investigated on polyamide- and sulfonated polyether sulfone-based NF membranes. A comprehensive characterization using field emission scanning electron microscopy, atomic force microscopy, Fourier transform infrared spectroscopy, Raman imaging, ellipsometry, contact angle measurements, and chemometric analysis was performed to establish relationships between membrane chemistry, morphology, multilayer formation, and separation performance.LBL modification significantly enhanced membrane selectivity and transport properties. For polyamide-based membranes, the best performance was obtained at 5.5 bilayers for PDADMAC/PSS and 6.5 bilayers for PAH/PSS, providing improved Na/Ca, Na/Mg, K/Ca, and K/Mg selectivity together with high water permeability. For sulfonated polyether sulfone membranes, lower polyelectrolyte concentrations (0.01 wt.%) produced more compact, homogeneous, and chemically stable multilayers than higher concentrations, demonstrating the importance of precise fabrication conditions.The thesis also explored thin-film composite and ultra-loose NF membranes incorporating metal-organic framework (MOF) nanoparticles. ZIF-8-NH₂ and UiO-66-NH₂ nanoparticles were synthesized and incorporated at different concentrations (0.075 and 0.15 wt.%) into the interlayer and active layer. Membrane performance strongly depended on MOF type, loading, and location within the membrane. A novel configuration combining UiO-66-NH₂ in the interlayer with ZIF-8-NH₂ in the active layer provided the best balance between permeability and ion separation.To summarize, this work demonstrates that LBL surface engineering and MOF incorporation are effective strategies for tailoring NF membrane properties. The modified membranes exhibited enhanced ion selectivity, permeability, and stability compared with commercial membranes while providing new insights into membrane structure–performance relationships. These findings contribute to the development of next-generation NF membranes for sustainable water treatment, hard water softening, and selective recovery of valuable ions, while providing a framework for future membrane design and characterization.
  

DOCTORAL DEGREE IN CIVIL ENGINEERING

• FARSHBAF, SIMA: Computational and Data-Driven Design of Additively Manufactured Auxetic Metamaterials for Enhanced Energy Absorption
  
  Author: FARSHBAF, SIMA
  Programme: DOCTORAL DEGREE IN CIVIL ENGINEERING
  Department: Barcelona School of Civil Engineering (ETSECCPB)
  Mode: Article-based thesis
  Deposit date: 29/09/2026
  Reading date: 30/11/2026
  Reading time: 11:00
  Reading place: Sala Zienkiewich (CIMNE) Building C1, UPC - Campus North Gran Capitan S/N 08034 Barcelona
  Thesis director: DIALAMISHABANKAREH SOLTANI, NARGES | CERVERA RUIZ, LUIS MIGUEL
  Thesis abstract: This doctoral thesis presents a computational-experimental framework for the analysis, design, and optimization of auxetic metamaterials with enhanced energy absorption performance. The proposed methodology combines experimental characterization, nonlinear finite element analysis, additive manufacturing, structural design, and machine learning to enable the efficient development of lightweight metamaterials with tailored mechanical properties.The thesis first investigates the mechanical behavior of conventional and auxetic cellular structures through experimental testing and validated nonlinear finite element simulations. The analyses provide insight into the influence of cellular topology, loading direction, and deformation mechanisms on the mechanical response, demonstrating the superior energy absorption capability of re-entrant auxetic structures over conventional honeycomb architectures.Building on these findings, novel geometrical modifications of the conventional re-entrant unit cell are proposed to enhance its auxetic response and energy absorption capacity while preserving its characteristic deformation mechanisms. The numerical predictions are experimentally validated, confirming the effectiveness of the proposed designs. Compared with the base design, the modifications achieve improvements of up to 1250\% in energy absorption capacity and 860\% in specific energy absorption.To accelerate the design process, a machine learning-based forward and inverse design framework is developed using datasets generated from validated finite element simulations. The proposed methodology enables efficient prediction of nonlinear mechanical responses and inverse identification of unit-cell geometries for both uniform and thickness-graded auxetic structures with prescribed mechanical performance. Furthermore, a physics-guided machine learning framework that combines a layer-wise response decomposition strategy with deep neural networks is developed to substantially reduce the computational cost of dataset generation while enhancing predictive performance. The proposed framework achieves prediction errors below 2.5\%, including in extrapolation scenarios, using only 10 training samples, compared with 2,145 samples required by the conventional deep neural network.Overall, this thesis establishes a framework integrating computational mechanics, experimental validation, structural optimization, and physics-guided machine learning for the efficient design of advanced auxetic metamaterials. The proposed methodologies provide practical tools for developing lightweight structures with tailored mechanical performance and enhanced energy absorption, while providing a methodological basis for extending data-driven design approaches to other metamaterial systems.
  

DOCTORAL DEGREE IN COMPUTATIONAL AND APPLIED PHYSICS

• BENÍTEZ COLOMINAS, POL: Modeling Lattice-Mediated Optoelectronic Response in Anharmonic Semiconductors: A First-Principles and Machine Learning Study
  
  Author: BENÍTEZ COLOMINAS, POL
  Programme: DOCTORAL DEGREE IN COMPUTATIONAL AND APPLIED PHYSICS
  Department: Department of Physics (FIS)
  Mode: Article-based thesis
  Deposit date: 15/10/2026
  Reading date: 12/11/2026
  Reading time: 10:00
  Reading place: Sala Polivalent, Edifici A, Escola d'Enginyeria de Barcelona Est (EEBE), Campus Diagonal Besos
  Thesis director: CAZORLA SILVA, CLAUDIO | SAUCEDO SILVA, EDGARDO ADEMAR
  Thesis abstract: The rapid development of next-generation energy technologies, such as photovoltaics, necessitates the discovery of materials with highly tunable and resilient optoelectronic properties. This dissertation investigates the fundamental role of lattice dynamics in phase stability and thermal renormalization of the optoelectronic properties of anharmonic materials exhibiting strong electron-phonon coupling, through an integrated approach combining first-principles quantum simulations and machine learning.The first part of this work analyzes in detail a family of materials known as antiperovskites, with chemical formula Ag3BC (B: S, Se; C: Cl, Br, I), that is fundamentally very interesting and highly relevant for energy and optoelectronic applications. In particular, by employing crystal structure prediction and free-energy calculations, we demonstrate that the cubic phase identified in experiments is dynamically stabilized at finite temperatures or represents kinetically trapped metastable states. Following this, we quantify a giant thermally induced band-gap renormalization in these systems, revealing that large-amplitude phonon displacements at room temperature can narrow the electronic band gap by more than 50%. This finding not only reconciles theory with experiment but also reveals a sensitive coupling between polar lattice distortions and electronic hybridization. We leverage this insight to propose a novel paradigm for “electro-optoelectronic” tuning, demonstrating that external electric fields can mimic thermal effects to provide non-chemical control over the band gap in centrosymmetric semiconductors.To scale these findings to more complex systems, the second part of the thesis focuses on the development of physics-informed artificial intelligence (AI) workflows. We show that graph neural networks trained on datasets generated via phonon-informed sampling, rather than random displacements, exhibit superior accuracy and generalizability. Through model interpretability analysis, we prove that these networks effectively learn the relative importance of specific chemical bonds in determining electronic responses. Finally, we implement an AI-workflow to evaluate the thermal response of complex solid solutions (Ag3SBr(x)I(1-x)), successfully simulating systems that were previously computationally prohibitively expensive using standard methods. Collectively, this research establishes that the explicit inclusion of lattice dynamics is essential for the accurate description of anharmonic materials and provides a robust, AI-accelerated framework for the “temperature-aware” design of functional semiconductors.
  
• DYHR, SØREN ISTVÁN ADORJÁN: Optimal Metrics in Odd-Dimensional Geometry: Reeb Dynamics and Applications in Fluid Mechanics
  
  Author: DYHR, SØREN ISTVÁN ADORJÁN
  Programme: DOCTORAL DEGREE IN COMPUTATIONAL AND APPLIED PHYSICS
  Department: Department of Physics (FIS)
  Mode: Normal
  Deposit date: 11/09/2026
  Reading date: 12/11/2026
  Reading time: 13:00
  Reading place: Sala de juntes Facultat de Matemàtiques i Estadística, Campus Sud - UPC
  Thesis director: MIRANDA GALCERÁN, EVA | GONZÁLEZ PRIETO, ÁNGEL
  Thesis abstract: Particles described by Hamiltonian mechanics tend to stay on odd-dimensional hypersurfaces of symplectic manifolds. In this thesis we study a generalization of the geometric structures appearing on such submanifolds. They all have a Reeb vector field and, in dimension three, coincide with stable Hamiltonian structures (SHS), a class which includes both cosymplectic and contact structures. We study two problems concerning metrics compatible with such structures.First we consider the Navier-Stokes equations, which describe viscous incompressible fluids. It turns out that Reeb vector fields of cosymplectic structures are stationary solutions for all compatible metrics. We create a cosymplectic plug from a construction by Cardona, Miranda, Peralta-Salas, and Presas, which realizes Turing machines as vector fields. Their construction was originally used in contact geometry to extend these vector fields to stationary solutions of the 3D Euler equations, which model ideal (inviscid) fluids. We extend their result by embedding the plug to obtain cosymplectic Reeb fields, and thus stationary solutions to the 3D Navier-Stokes equations, capable of simulating any Turing machine.Second we study an energy functional on the space of compatible metrics, originally considered for contact manifolds by Blair as well as by Chern and Hamilton. The metrics of minimal energy for contact manifolds in dimension three have been classified. We generalize the functional and the classification to three-dimensional SHSs and find that positive energy critical metrics correspond to the algebraic Anosov flows.Finally we obtain a similar result in higher dimension when there exist certain first integrals satisfying a non-degeneracy condition. We also classify a family of five-dimensional examples of critical metrics, including some not satisfying the non-degeneracy condition.
  

DOCTORAL DEGREE IN COMPUTER ARCHITECTURE

• CARO ROCA, MARTÍ: Efficient Tailoring of AI-based Solutions for Safety-Critical Systems
  
  Author: CARO ROCA, MARTÍ
  Programme: DOCTORAL DEGREE IN COMPUTER ARCHITECTURE
  Department: Department of Computer Architecture (DAC)
  Mode: Normal
  Deposit date: 10/09/2026
  Reading date: 19/11/2026
  Reading time: 10:15
  Reading place: C6-E101
  Thesis director: ABELLA FERRER, JAIME
  Thesis abstract: The increasing performance of computing platforms enables the use of larger Deep Neural Networks (DNNs) that achieve the accuracy needed for real-time applications such as Camera-based Object Detection (CBOD) in Autonomous Driving (AD). While high inference accuracy is mandatory for DNNs in Critical Real-Time Embedded Systems (CRTES), memory bandwidth and energy consumption are also critical factors, often constraining the achievable performance and determining whether those DNNs meet application requirements (e.g., frames per second). Therefore, some highly accurate but large DNN models must be optimized to operate under these constraints.Moreover, automotive applications with safety requirements must adhere to standards such as ISO 26262, which impose the use of diverse redundancy for the highest integrity level (ASIL D) to mitigate common-cause faults — random hardware faults that can potentially affect redundant components similarly and remain undetected. Dual Modular Redundancy (DMR), also referred to as Dual-Core LockStep (DCLS) for CPUs, and Triple Modular Redundancy (TMR), often combined with some form of diversity, are commonly used in safety-critical systems to provide fault detection and tolerance. However, efficiently realizing diverse redundancy for DNN-based safety-related tasks, such as object detection, remains an open challenge, as these applications can leverage semantic redundancy rather than bit-level equivalence.In this thesis, we first perform an at-scale evaluation of several approximation domains to optimize DNN models. In particular, we focus on CBOD models, which are essential in AD to detect objects surrounding the vehicle. We study several techniques independently and jointly — including reduced-precision arithmetic, approximate computing, and low-voltage operation — to assess how they can be combined to significantly reduce the area, delay, and energy of AI systems. We further extend this analysis by studying the impact of weight clustering for different arithmetic representations, including floating-point and integer formats, providing insights to guide the development of low-power CBOD accelerators with negligible impact on accuracy.Next, we design an efficient diverse redundancy scheme tailored to AI applications by leveraging reduced precision. The proposed scheme operates on reduced-precision arithmetic obtained by dropping selected bits from the fetched data of the main accelerator. This enables mitigation of random hardware faults while reducing the delay, area, and energy of the redundant accelerator compared to traditional DCLS implementations, without noticeable accuracy loss.We also propose a diverse redundancy technique with a similar cost to DCLS-like implementations but with a significant accuracy increase, achieved by mitigating AI model inaccuracies and avoiding the need for specific hardware support. This scheme applies input transformations (e.g., horizontal flip or Gaussian noise) and uses merging algorithms to fuse detections from redundant components. We further evaluate its ability to mitigate adversarial attacks, allowing a single mechanism to simultaneously address random hardware faults, DNN model inaccuracies, and adversarial attacks. Finally, we show that the scheme can be generalized to other AI models and tasks, confirming that the observed conclusions and trends hold across different AI modalities.Overall, this thesis studies approximation techniques that can significantly optimize AI models for CRTES and proposes two enhanced diverse redundancy techniques: one targeting energy efficiency and another focusing on improving the accuracy of the diverse redundant system with negligible overhead.
  
• NIU, YUYAO: Hardware-Aware Algorithms for Efficient Graph Traversal on High-End Parallel Architectures
  
  Author: NIU, YUYAO
  Programme: DOCTORAL DEGREE IN COMPUTER ARCHITECTURE
  Department: Department of Computer Architecture (DAC)
  Mode: Normal
  Deposit date: 09/09/2026
  Reading date: 06/11/2026
  Reading time: 11:00
  Reading place: Sala d'Actes B6
  Thesis director: CASAS GUIX, MARC
  Thesis abstract: Graph traversal is a fundamental building block in scientific computing, data analytics, and artificial intelligence. Among traversal methods, Breadth First Search (BFS) and Depth First Search (DFS) are two core primitives for higher-level graph algorithms. As graph-structured data continues to grow, efficient traversal increasingly depends on the ability to exploit high-end parallel architectures. However, modern architectures rely on structured and throughput-oriented execution engines, such as long vector processors, dense matrix accelerators, and massively parallel GPUs, whereas graph traversal exhibits irregular memory accesses, data-dependent control flow, and dynamic workloads.This dissertation argues that the performance gap between graph traversal and modern hardware arises from three major mismatches: a structural mismatch between regular execution units and irregular graph structure, a computational mismatch between arithmetic-oriented acceleration and relation-driven traversal, and an architectural mismatch between throughput-oriented parallelism and dependency-driven traversal behavior. To address these challenges, the dissertation develops a hardware-aware design paradigm in which graph traversal algorithms are reformulated according to the characteristics and constraints of the target architecture.The first contribution is ALBBA, an algebraic BFS algorithm for long vector architectures. ALBBA reformulates direction-optimized BFS using a vector-friendly sparse representation, vectorized search, and a two-level bypass mechanism. These techniques allow the key algebraic BFS operations, sparse matrix-vector multiplication (SpMV) and sparse matrix-sparse vector multiplication (SpMSpV), to better exploit long vector execution while reducing unnecessary memory accesses and redundant computation. This shows that structural irregularity can be mitigated through hardware-aware data layout and execution design.The second contribution is BerryBees, a BFS framework for Tensor Core-based GPUs. BerryBees introduces the Binarized Row Slice format, which uses bit-level representations to support both SpMV and SpMSpV within a unified data layout while reducing memory overhead. It further employs a warp-level algorithm that leverages Tensor Core matrix-multiply-accumulate instructions to accelerate traversal. This shows that relation-driven graph computation can be reorganized to exploit hardware originally designed for dense matrix operations.The third contribution is DiggerBees, a parallel DFS algorithm for massively parallel GPUs. DiggerBees introduces a two-level stack structure aligned with the GPU memory hierarchy, warp-level DFS execution within thread blocks, and hierarchical work stealing across warps and blocks. These techniques expose scalable parallelism for dependency-driven traversal and improve load balance across the GPU. This shows that DFS, despite its inherent dependencies, can be effectively adapted to throughput-oriented parallel hardware.Together, these contributions demonstrate that BFS and DFS can be effectively accelerated when their data representations, computational formulations, and execution structures are designed with architectural constraints as first-class considerations. More broadly, this dissertation establishes hardware-aware reformulation as a general methodology for rethinking irregular graph algorithms on increasingly specialized parallel hardware.
  

DOCTORAL DEGREE IN COMPUTING

• PONS RECASENS, GERARD: Evolving Knowledge Graphs: Continuous Knowledge Acquisition, Adaptation, and Exploitation
  
  Author: PONS RECASENS, GERARD
  Programme: DOCTORAL DEGREE IN COMPUTING
  Department: Department of Computer Science (CS)
  Mode: Normal
  Deposit date: 01/10/2026
  Reading date: 11/11/2026
  Reading time: 15:00
  Reading place: Sala d'Actes FIB - Edifici B6 Campus Nord
  Thesis director: QUERALT CALAFAT, ANNA | BILALLI, BESIM
  Thesis abstract: Modern systems continuously generate vast amounts of heterogeneous data. While this data contains valuable information, its full potential remains difficult to realize when it is treated as a collection of isolated observations lacking meaning and context. Unlocking its full value therefore requires enriching data with explicit semantics that capture concepts, relationships, and contextual information, transforming isolated observations into structured knowledge. This thesis argues that such semantic enrichment constitutes a fundamental step towards building more effective intelligent systems. In this context, Knowledge Graphs (KGs) provide a natural framework for representing and organizing this knowledge, while Knowledge Graph Embeddings (KGEs) transform it into latent representations that can be flexibly exploited by modern Machine Learning methods to address evolving analytical needs.As knowledge is not static, realizing this vision requires addressing challenges that span its entire life cycle. First, data originating from heterogeneous and unstructured sources must be incorporated into a continuously evolving knowledge structure, where it is enriched with semantics. Second, as this knowledge structure evolves, its latent representations must be continuously updated while preserving previously acquired knowledge. Finally, the evolving knowledge structure and its learned representations must be effectively exploited to support adaptive behavior in intelligent systems.This thesis addresses these challenges through the framework of evolving KGs. For knowledge acquisition from unstructured sources, it focuses on textual information and proposes a schema-aware entity disambiguation method that combines Large Language Models (LLMs) with the hierarchical structure of KGs, improving adaptability across domains without requiring model retraining. To maintain latent representations over time, it investigates the continual adaptation of KGEs. Within this scope, it identifies a previously overlooked phenomenon that contributes to the degradation of previously acquired knowledge and proposes an adjusted evaluation protocol to accurately capture its effects. Additionally, it introduces an initialization strategy for newly added embeddings that enhances knowledge acquisition while mitigating this performance degradation. Finally, for knowledge exploitation, it demonstrates how continuously evolving KGs and their learned representations can support adaptive recommendation and intent-aware assistance in experiment-driven data analytics environments. Together, these advancements contribute towards enabling evolving KGs to continuously acquire, adapt, and exploit knowledge.
  

DOCTORAL DEGREE IN CONSTRUCTION ENGINEERING

• ALAHMAD, AHMAD: Joint State and Parameter Estimation for Structural System Identification: A Unified Statistical Framework across Static and Modal Data
  
  Author: ALAHMAD, AHMAD
  Programme: DOCTORAL DEGREE IN CONSTRUCTION ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 29/09/2026
  Reading date: 10/11/2026
  Reading time: 11:15
  Reading place: UPC Campus Nord, Carrer de Jordi Girona, 1-3, 08034 BarcelonaEDIFICI C1- AULA 116
  Thesis director: TURMO CODERQUE, JOSE | MINGUEZ SOLANA, ROBERTO
  Thesis abstract: Assessing the condition of aging civil infrastructure rests on Structural System Identification (SSI): recovering physical parameters such as element stiffnesses and boundary conditions by reconciling a numerical model with measured responses. Every ingredient of this inverse problem is uncertain: measurements carry noise, loads and supports are imperfectly known, and the model idealizes the real structure. A trustworthy result must therefore do more than return a point estimate, yet least-squares estimation, sensitivity analysis, probabilistic uncertainty treatment, and observability theory have largely been pursued separately in structural identification rather than wired into a single estimator.The power systems community faced this challenge earlier. Since Static State Estimation (SSE) was formulated in the 1970s, it has built around the weighted least-squares (WLS) estimator an integrated statistical apparatus refined over more than five decades of operational use: pre-estimation observability analysis, covariance-based uncertainty quantification, bad-data detection through the largest normalized residual, and χ² testing of model adequacy. This thesis adapts that apparatus to structural identification and extends it from state-only to joint state and parameter estimation, with observability as a pre-estimation guard and the residual diagnostics as post-estimation guards, in a single WLS interface accepting static and operational-modal measurements.The methodology comprises five developments, validated in two experimental campaigns. A Jacobian-based observability criterion fixes parameter identifiability from the rank of the measurement Jacobian, replacing recursive symbolic methods at a fraction of their cost. The estimator is then equipped, for the first time in SSI to the author's knowledge, with covariance-based confidence intervals, χ² model-adequacy testing, and normalized-residual bad-data detection. Model-form error is treated directly: unknown loads and boundary conditions become estimable variables, removing the systematic bias they would induce, while principal-component analysis of the residuals localizes what remains. A resolution-matrix conservation diagnostic certifies in advance which property groups admit reliable integrated-stiffness recovery. Finally, static and modal residuals combined over a shared parameter vector localize a stiffness change as accurately as either modality alone, while an additive Fisher-information decomposition breaks the mass-stiffness degeneracy modal data cannot resolve in isolation.The framework is validated on a common laboratory specimen instrumented with low-cost MEMS sensors. A static campaign under an automated multi-load-case protocol confirms that the χ²-calibrated noise model transfers unchanged to an independent test, that sparse rotation measurements under diverse loading resolve a spatially detailed stiffness profile, and that the conservation integral recovers a reinforced section's stiffness to within a fraction of a percent of theory. A second campaign re-instruments the same specimen for output-only modal testing; with only the modal residual rows added, the same solver and diagnostics recover the same reinforcement, and the static and modal estimates agree with the theoretical reference to within about two percentage points. It also maps a limit of the self-diagnosis: an unmodeled support compliance, nearly collinear with the boundary stiffness, biases the boundary parameters without a signature in the diagnostics, surfacing only when the supports are modeled explicitly.
  
• BASAR, IMREN: Sustainable Repair Mortars for Historic Masonry: Design, Characterization and Performance Assessment of Natural and Recycled Materials
  
  Author: BASAR, IMREN
  Programme: DOCTORAL DEGREE IN CONSTRUCTION ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 29/09/2026
  Reading date: 10/11/2026
  Reading time: 12:00
  Reading place: ETSECCPB, Sala C2-212 (C/Jordi Girona 1-3, mòdul C2, Campus Nord, Barcelona).
  Thesis director: FERNANDEZ CARRASCO, LUCIA | TORRENS MARTÍN, DAVID
  Thesis abstract: This thesis presents a comparative and conservation-oriented investigation of lime-based mortars employed in architectural heritage, with particular emphasis on material compatibility, durability, and long-term performance. Historical mortars sampled from two representative sites in Catalonia: the Roman archaeological site of Altafulla and the Edifici de l’Hivernacle in Barcelona, were studied alongside commercial restoration mortars and laboratory-designed lime-based formulations in order to evaluate technological choices, performance limits, and conservation suitability.A comprehensive multi-analytical methodology was adopted, integrating mineralogical, granulometric, mechanical, physical, and visual characterization techniques. This approach enabled the assessment of binder–aggregate relationships, microstructural features, mechanical response, and aesthetic behaviour, with special attention to parameters relevant to heritage conservation practice. The characterization of historical mortars revealed distinct technological strategies in raw material selection, aggregate grading, and binder composition, closely linked to the functional requirements and environmental exposure conditions of each site.Comparative analysis demonstrated that standardized commercial repair mortars, despite exhibiting higher mechanical strength, do not necessarily ensure improved compatibility with historic masonry substrates. In contrast, laboratory-prepared lime mortars, particularly CL90 air-lime-based systems, showed greater adaptability in balancing mechanical performance and deformability when curing conditions and environmental exposure were carefully controlled.Durability-related conditioning tests highlighted the critical influence of carbonation regime, moisture interaction, salt presence, and thermal cycling on mortar behaviour over time, underscoring the necessity of long-term evaluation beyond early-age mechanical performance. In situ qualitative observations conducted at the Monestir de Sant Joan les Fonts further confirmed satisfactory workability, adhesion, and visual integration of both reference and PDMS-modified lime mortars during the initial post-application period.Taken together, the results support an integrated and context-sensitive approach to the design and assessment of compatible lime-based repair mortars, providing experimental evidence that emphasizes the importance of aesthetic or integration considerations, material compatibility, durability, and environmental interaction in the conservation of architectural heritage.Keywords: Lime-based mortars, heritage conservation, material compatibility, durability, long-term performance, carbonation, CL90 air lime, repair mortars, microstructure, environmental exposure, thermal cycling, salt interaction, aesthetic considerations, architectural heritage. 
  
• VERGÉS EIRAS, ROGER: Artificial Intelligence for Circular and Sustainable Product Decision-Support in Construction
  
  Author: VERGÉS EIRAS, ROGER
  Programme: DOCTORAL DEGREE IN CONSTRUCTION ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Article-based thesis
  Deposit date: 01/10/2026
  Reading date: 06/11/2026
  Reading time: 11:00
  Reading place: (ESEIAAT): C/ Colom, 1-11, 08222 Terrassa (Barcelona). Edifici TR5 AULA 2.0b
  Thesis director: FORCADA MATHEU, NURIA | GASPAR FÀBREGAS, KÀTIA
  Thesis abstract: Construction's environmental footprint and the regulatory targets now set to reduce it have made the sector's transition toward sustainability and circularity an urgent priority. Decisions about which products and materials to specify, made while a building is still only partially defined, carry the most significant influence over its material and environmental trajectory. Yet, the tools developed to support such decisions, life cycle assessment chief among them, presuppose levels of data completeness that early design is structurally unable to provide. This gap is often bridged by cost and availability alone; if sustainability is considered at all, product selection typically involves expert judgement applied on a case-by-case basis, a response that scales poorly and leaves little traceable reasoning behind. This mismatch between decision-making conditions and tool assumptions defines the problem this thesis addresses.This thesis develops an integrated, end-to-end pipeline that pursues four objectives in sequence, realised through five interconnected contributions reported as journal articles: a product-level decision framework establishing the indicators, product information model, and relative evaluation logic against which construction products are compared under early-design uncertainty; a large-scale empirical characterisation of how environmental and circularity information is reported across construction product declarations, together with strategies for handling heterogeneous, inconsistent, and missing data; a human-in-the-loop screening agent that filters candidates against regulatory and project-specific requirements; an inference model that estimates product-relevant attributes where declared data is absent; and a context-adaptive recommender model that completes the available evidence and adapts to the stakeholder priorities and application requirements of each decision context.The pipeline establishes several findings that extend beyond the individual contributions. Relative comparison, applied consistently across environmental, circularity, economic, and performance dimensions, emerges as the evaluation logic early design can sustain: it depends only on the evidence assembled across alternatives being consistent, so a partial or uneven data landscape does not invalidate the comparison the way it would an absolute method such as life cycle assessment. Incomplete product data is treated as an expected feature of early design, characterised empirically, partially resolved through inference, and carried forward as a legible signal in the output, so the comparison remains trustworthy when product choices are still open. The resulting recommendation system is relational and learned directly from data, adapting to each decision context. Validation with concrete as case-study product category demonstrates that extraction quality, compliance outcomes, and inferred attributes propagate coherently through to the final recommendation, a property none of the five contributions could establish in isolation.For designers and procurement teams, the implication is that environmental and circularity performance become specification criteria from the outset, before cost and availability alone have narrowed the field, because product data has been structured into a standardised, machine-readable form. When multiplied across the volume of decisions the sector makes, that shift turns individual product choices into measurable progress on both fronts, without cost or availability having to give ground.
  

DOCTORAL DEGREE IN EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

• CUESTA ARRILLAGA, JOKIN: Physics-integrated artificial intelligence prognostics for wind turbine components under uncertainty and data scarcity
  
  Author: CUESTA ARRILLAGA, JOKIN
  Programme: DOCTORAL DEGREE IN EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 11/09/2026
  Reading date: 20/11/2026
  Reading time: 10:00
  Reading place: EEBE. CAMPUS BESOSEdifici ASala PolivalentAvinguda d'Eduard Maristany, 16, 08019 Barcelona / Sant Adrià de Besòs
  Thesis director: VIDAL SEGUI, YOLANDA | LETURIONDO ZUBIZARRETA, URKO | POZO MONTERO, FRANCESC
  Thesis abstract: Wind energy plays a central role in the global transition towards sustainable energy systems. However, the operation and maintenance costs associated with wind turbines remain a major challenge, particularly due to unexpected component failures and the difficulty of accurately predicting their remaining useful life. Prognostics and health management has emerged as a key enabler of predictive maintenance strategies, yet current prognostic approaches often face limitations related to data scarcity and quality, limited observability of degradation, lack of physical consistency, and inadequate uncertainty quantification.This thesis addresses these challenges by developing adaptive physics-informed prognostic methodologies for wind turbine components operating under uncertainty and with scarce degradation data. A unified uncertainty-aware prognostic framework is proposed that integrates physical knowledge into data-driven models based on the observability level of the degradation process. For directly observable degradation scenarios, physics-informed neural network architectures based on time-series transformers are developed by embedding degradation laws into the learning process. For indirectly observable degradation scenarios, where degradation cannot be measured directly, a physics-guided framework is introduced to construct health indicators from operational data through residual-based representations of healthy system behavior.The proposed methodologies are validated through a two-level experimental strategy. First, a dedicated laboratory testbed is designed to generate controlled, run-to-failure wear data. Second, the framework is assessed using real wind turbine SCADA data from gearbox degradation cases to demonstrate its applicability in operational environments. In both scenarios, multiple uncertainty quantification techniques are investigated and benchmarked to characterize predictive confidence and support maintenance decision-making.The results demonstrate that integrating physical knowledge into advanced deep learning architectures improves degradation forecasting accuracy, robustness, and remaining useful life estimation performance compared with purely data-driven approaches. Furthermore, the proposed uncertainty-aware framework provides reliable confidence estimates, enhancing the interpretability and operational usefulness of prognostic predictions. The findings contribute to the development of trustworthy hybrid prognostic systems for wind energy applications and maintenance timing, supporting wind-farm availability management and the dependable operation of renewable energy assets.
  
• WANG, SHUN: Artificial Intelligence for Wind Turbine Condition Monitoring and Fault Diagnosis Towards Predictive Maintenance
  
  Author: WANG, SHUN
  Programme: DOCTORAL DEGREE IN EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Article-based thesis
  Deposit date: 21/09/2026
  Reading date: 23/11/2026
  Reading time: 10:00
  Reading place: EEBE. CAMPUS BESOSEdifici ASala PolivalentAvinguda d'Eduard Maristany, 16, 08019 Barcelona / Sant Adrià de Besòs
  Thesis director: POZO MONTERO, FRANCESC | VIDAL SEGUI, YOLANDA
  Thesis abstract: The continued expansion of wind energy depends on the reliable operation of large, complex, and geographically dispersed turbines. Late fault detection can cause secondary damage, unplanned downtime, and costly maintenance. Artificial intelligence can learn condition patterns from turbine measurements, but its practical use is limited by scarce fault labels, changing operating conditions, noise, and interactions among variables.This doctoral thesis develops and validates physically informed artificial intelligence strategies for wind turbine predictive maintenance through three information routes: vibration monitoring, supervisory control and data acquisition (SCADA) monitoring, and thermal-image fault diagnosis. The research combines unsupervised representation learning, healthy-data-based anomaly detection, normal behavior modeling, and machine-learning classification. Vibration and SCADA data are used to detect departures from healthy operation, whereas thermal images distinguish labeled electrical equipment conditions. The central question is how representations, references, and inputs can preserve health-related information and support reliable detection or diagnosis. The vibration research progresses from component-level degradation assessment to system-level monitoring and robustness under variable speed. Across four public bearing run-to-failure datasets, the logarithmic envelope spectrum is combined with a variational autoencoder trained only on healthy data to produce an interpretable health index for early fault detection and degradation tracking. For vibration data from an operational research wind turbine, a variational autoencoder learns healthy feature distributions, while exponentially weighted moving average monitoring stabilizes the reconstruction-error indicator for detecting documented pitch faults, icing events, and aerodynamic imbalances. In controlled gearbox experiments, a speed-invariant autoencoder integrates order-domain alignment, cross-condition interpolation, and speed-conditioned reconstruction to obtain stable and threshold-consistent health indicators across the evaluated steady, variable, and unseen speeds. The SCADA-based research examines how input design affects the early-fault sensitivity of normal behavior models. Conventional models can conceal early degradation when highly correlated internal inputs carry the same fault influence as the monitored target. An exogenous-variable virtual sensor is therefore constructed using ambient temperature and wind speed, enabling a model to learn normal drivetrain thermal behavior from inputs external to component health. Across three gearbox- and generator-bearing fault cases from turbines of different manufacturers, the method provides warnings up to 93 days before maintenance and no false alarms during separately assessed healthy periods. The thermal-image research develops a non-contact diagnostic framework for electrical equipment. Modified multiscale two-dimensional color distribution entropy is introduced as a compact feature set extracted directly from pseudocolor thermograms before machine learning classification. In the evaluated induction-motor and power-transformer experiments, diagnostic accuracy exceeds 95\%, and the method performs better under the tested noise and limited-sample conditions than the compared grayscale entropy formulations.Within the evaluated datasets and experimental conditions, the vibration studies demonstrate early degradation assessment and system-level fault detection; the SCADA study provides earlier warnings for the documented drivetrain bearing cases; and the thermal-image study improves fault classification for the evaluated electrical equipment. Together, these results provide proof of principle for earlier and more informative condition assessment while identifying the fleet-level and field validation required before routine integration into wind turbine maintenance decisions.
  

DOCTORAL DEGREE IN ELECTRICAL ENGINEERING

• ALI, SALMAN: ANALYSIS OF THE INTEGRATION OF RENEWABLE ENERGY AND ENERGY STORAGE SYSTEMS IN SERIES-CASCADED MICROGRIDS
  
  Author: ALI, SALMAN
  Programme: DOCTORAL DEGREE IN ELECTRICAL ENGINEERING
  Department: Department of Electrical Engineering (DEE)
  Mode: Normal
  Deposit date: 29/09/2026
  Reading date: 05/11/2026
  Reading time: 10:00
  Reading place: ESEIAAT – TR5 – Expo Sala
  Thesis director: BOGARRA RODRIGUEZ, SANTIAGO | CORCOLES LOPEZ, FELIPE
  Thesis abstract: The increasing integration of renewable energy sources has transformed power distribution networks into decentralized energy systems. Distributed generation (DG) sources, including solar photovoltaic (PV), wind, fuel cells, hydropower, and other renewable resources, enable localized power generation and reduce dependence on centralized infrastructure. Microgrids (MGs) have emerged as a solution for integrating DG units. MG structures are widely adopted due to their control and protection characteristics. However, the low-voltage output of DG sources requires boost conversion stages, increasing system complexity and losses. Consequently, series-cascaded microgrids (SCMGs) have attracted attention.Despite their advantages, SCMGs face challenges in power sharing, voltage regulation, and integration of non-dispatchable DG (ND-DG) sources. Existing power routing controls generally assume ideal DC sources, limiting their applicability under renewable intermittency. Conventional approaches such as fundamental voltage reconstruction and third-harmonic voltage injection exhibit limited power sharing capability, constrained DC-link voltage margins, overmodulation risks, and degraded dynamic performance. To address these limitations, this thesis investigates an SCMG topology integrating ND-DG units and battery storage systems. A multi-power-carriers based control strategy incorporates third-harmonic current injection alongside fundamental component to enable inter-module power routing under reduced or zero ND-DG output. The proposed approach maintains desired multilevel output voltage while satisfying voltage and modulation constraints. In addition, proportional DC-side power utilization coordinates battery charging and discharging, supporting voltage stability and power sharing. Although the proposed methodology extends the power routing range, injected third-harmonic current appears in the grid current, affecting grid compatibility.To overcome this limitation, the SCMG architecture is extended to a delta-connected configuration. In this topology, third-harmonic current circulates within cascaded clusters, enabling inter-module power routing without propagating harmonics to the grid. Zero-sequence current injection is employed for inter-cluster power routing and symmetrical three-phase currents. The proposed framework determines optimal third-harmonic current, third-harmonic voltage, and zero-sequence injection, together with proportional DC-side power utilization control. Under adverse operating conditions, the fundamental component governs grid-compliant power exchange, the third-harmonic component enhances inter-module power redistribution, and the zero-sequence component ensures inter-cluster balancing.The proposed methodology is validated through MATLAB/Simulink simulations, hardware-in-the-loop experiments, and a laboratory-scale prototype. The results confirm stable multilevel voltage generation, effective DC-link stabilization, and robust inter-module and inter-cluster power routing. Overall, this thesis advances the state-of-the-art in SCMGs by introducing a scalable control framework that enables reliable operation under realistic microgrid conditions and supports renewable energy and energy storage integration in future smart power systems.
  

DOCTORAL DEGREE IN ELECTRONIC ENGINEERING

• HUSSAIN, TAUSEEF: Metamaterial-inspired Flexible Microwave Resonant Structures for RFID Sensing and Shielding
  
  Author: HUSSAIN, TAUSEEF
  Programme: DOCTORAL DEGREE IN ELECTRONIC ENGINEERING
  Department: Department of Electronic Engineering (EEL)
  Mode: Article-based thesis
  Deposit date: 20/07/2026
  Reading date: 05/11/2026
  Reading time: 15:30
  Reading place: Sala de Conferències TR1 – ESEIAAT, UPChttps://meet.google.com/fkh-mqus-zbt
  Thesis director: GIL GALI, IGNACIO | FERNANDEZ GARCIA, RAUL
  Thesis abstract: Metamaterials are artificially engineered electromagnetic structures that enable wave–matter interactions beyond those achievable with natural materials. When implemented in flexible and textile-compatible forms, metamaterial-inspired resonant structures offer unique advantages for wearable technologies, including conformability, lightweight construction, and tunable electromagnetic responses. These characteristics make them particularly attractive for wearable radio-frequency identification (RFID) sensing platforms, where sensitivity, robustness, and user safety must be addressed simultaneously. This thesis investigates the design and integration of metamaterial-inspired flexible resonant structures with RFID systems to enable sensors with enhanced performance and improved electromagnetic safety.Conventional RFID-based sensors typically rely on antenna deformation or substrate perturbation alone to transduce physical or biochemical stimuli. While effective in controlled environments, these approaches exhibit limited sensitivity and strong dependence on environmental loading in wearable scenarios. At ultra-high-frequency (UHF) bands, proximity to the human body causes impedance detuning and reduced read range reliability, whereas at microwave frequencies, increased absorption in biological tissue raises specific absorption rate (SAR) concerns. These limitations motivate the use of engineered resonant structures that offer enhanced field confinement and improved control over antenna–environment interactions.This thesis first investigates resonator-assisted passive UHF RFID sensing for biomechanical and biochemical monitoring. Electromagnetic coupling between printed UHF RFID tag antennas and compact resonant elements is exploited to enhance sensitivity beyond conventional antenna-based sensing. Meander-line resonators are integrated with RFID tags to enable displacement sensing, demonstrating reliable respiration and knee-flexion monitoring. The coupling approach is further extended to biochemical sensing through the integration of a capacitively coupled complementary split-ring resonator (CSRR) with an H-slot RFID antenna, enabling passive wireless monitoring of glucose concentration via read-range modulation.Subsequently, the thesis addresses electromagnetic stability and user safety at microwave RFID frequencies, where shorter wavelengths enable compact resonator arrays to be realized as metasurfaces. A breathable ground-grid artificial magnetic conductor (GG-AMC) is developed to mitigate body-induced detuning, suppress backward radiation, and significantly reduce SAR while preserving wearer comfort. In addition, near-field coupled metamaterial absorbers are introduced as dual-function structures that simultaneously provide electromagnetic shielding and strain-sensitive resonance modulation, enabling respiration monitoring at microwave frequencies. The core scientific contributions of this research are documented in five JCR-indexed journal articles and two peer-reviewed IEEE conference publications, and the thesis is presented as a compendium of articles.Keywords – RFID sensing, wearable sensors, microwave resonators, metasurfaces, biomechanical monitoring, biochemical sensing, artificial magnetic conductor, metamaterial absorber, electromagnetic shielding.
  
• VAQUERO GALLARDO, NOELIA: Contribution to Advanced Electrical Sensing and Predictive Control for Enhanced Safety in Cryolipolysis Applications
  
  Author: VAQUERO GALLARDO, NOELIA
  Programme: DOCTORAL DEGREE IN ELECTRONIC ENGINEERING
  Department: Department of Electronic Engineering (EEL)
  Mode: Article-based thesis
  Deposit date: 01/10/2026
  Reading date: 05/11/2026
  Reading time: 11:00
  Reading place: Aula A0.02, Edifici
  Thesis director: MARTINEZ GARCIA, HERMINIO | MILLÁN BLASCO, OLIVER
  Thesis abstract: Active medical devices based on energy application, such as radiofrequency, cryotherapy, or electrosurgery, are widely used due to their therapeutic effectiveness in non-invasive treatments. However, one of their main side effects is the occurrence of thermal injuries and skin burns, associated with inadequate energy delivery or insufficient monitoring of tissue status. In clinical practice, many of these devices do not integrate advanced sensing and control techniques, operating with fixed parameters regardless of patient-specific characteristics, which leads to significant variability in therapeutic outcomes. Therefore, monitoring mechanisms are required to assess tissue status and dynamically adjust treatment parameters, ensuring both efficacy and safety.This Ph.D. Thesis addresses this challenge through a multidisciplinary approach based on fractional electrical modeling, early detection of tissue freezing, and adaptive process control derived from the electrical and dielectric variations of biological tissues. First, a modeling framework based on fractional-order systems is established to describe the electrical behaviour of technical coatings and biological tissues. Using Cole-family models fitted to experimental impedance data, it is demonstrated that fractional models accurately reproduce the complex electrical response while using a reduced number of parameters, thereby simplifying instrumentation and facilitating the miniaturization of embedded systems. Building on this framework, the use of bioimpedance and capacitance is explored as tools for the early detection of tissue freezing, providing alternatives to methods based solely on temperature. Bioimpedance enables freezing detection by measuring the increase in resistance and the decrease in conductance that occur when water within the tissue turns into ice, whereas capacitance reflects the reduction in dielectric constant associated with ice crystal formation, thereby identifying early changes in water structure. Finally, a predictive model based on Random Forest is developed, integrating bioimpedance, capacitance, and temperature signals to anticipate freezing events.The results of this Ph.D. Thesis show that predictive detection occurs, on average, 16 seconds earlier than capacitive detection and 63 seconds earlier than temperature–bioimpedance-based detection. This model enables the implementation of a real-time adaptive control strategy over the cooling system, reinforcing operational safety. In summary, this Ph.D. Thesis demonstrates that the electrical properties of tissues allow freezing events to be identified before conventional temperature indicators, establishing a clinically relevant framework for intelligent, non-invasive cryotherapy devices that improve efficacy, safety, and regulatory compliance.
  

DOCTORAL DEGREE IN ENVIRONMENTAL ENGINEERING

• LAGE JULIÀ, ARTAI: Process engineering development strategies for the enhancement of PHB production by cyanobacteria.
  
  Author: LAGE JULIÀ, ARTAI
  Programme: DOCTORAL DEGREE IN ENVIRONMENTAL ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 12/10/2026
  Reading date: 18/11/2026
  Reading time: 10:30
  Reading place: Place: ETSECCPBUPC, Campus NordBuilding C2. Classroom: 212C/Jordi Girona, 1-308034 Barcelona
  Thesis director: GARCIA SERRANO, JOAN | GONZALEZ FLO, EVA
  Thesis abstract: Conventional petroleum-based plastics cause severe, lasting environmental harm. Polyhydroxybutyrate (PHB) is a promising fully biodegradable, biobased alternative, as cyanobacteria can synthesize it photoautotrophically from sunlight and CO2, bypassing the costly feedstocks that make bioplastic production economically uncompetitive. Yet several bottlenecks prevent cyanobacterial PHB reaching industrial scale. This thesis addresses them via an integrated strategy: strain selection and microbiome assembly, high-throughput methodology development and parameter optimization, and engineering a decoupled continuous production architecture.Five wild-type cyanobacterial strains were isolated from natural aquatic environments via single-cell isolation, yielding stable populations with markedly divergent PHB accumulation despite high genetic similarity within species. Two Synechocystis sp. and one Leptolyngbya sp. isolate emerged as the top producers, while Synechococcus sp. and a second Leptolyngbya sp. isolate showed negligible accumulation. Secondary effluent was detrimental to accumulation, as residual nutrients interfered with induction, but remained viable for biomass growth.The two best strains (Synechocystis sp. AP, Leptolyngbya sp. R) were evaluated as monocultures, while an artificial microbiome combining Synechocystis sp. AP with the low-producing Synechococcus sp. AP tested community-level synergies. Co-cultivation enhanced PHB performance in both species beyond monoculture levels, and the microbiome was more robust under non-sterile conditions, sustaining stable, contamination-free operation across consecutive cycles. This microbiome showed functional redundancy: under divergent conditions (low phosphorus with uncontrolled pH above 10, versus nutrient-replete conditions with strict pH control at ~8.25), relative population abundances shifted considerably, yet overall PHB potential remained intact. Differences in PHB content were attributable primarily to operational parameters, not population composition: strict pH control and sufficient nutrients during growth enabled up to 36.3% dcw, versus only 17% when pH was uncontrolled.To enable rapid screening, a fluorometric platform based on Nile Blue staining was developed and validated, enabling 124 conditions to be evaluated without the demands of chromatographic quantification. Nitrogen depletion was confirmed as the primary, continuous induction trigger, while phosphorus acted as a threshold switch. Strict pH control (~8.25), dark anoxic conditions, and a 600 mg/L acetate dose were identified as optimal, and validated at photobioreactor scale, confirming PHB synthesis directly from acetate without a glycogen pre-accumulation step.Building on this microbiome, expanded with Leptolyngbya sp. R into a three-strain community, a novel two-stage reactor decoupling photoautotrophic biomass growth from dark, anoxic PHB accumulation was designed and validated for 125 days under non-sterile conditions. Eliminating lags inherent to sequential single-reactor systems, this delivered ~23 mgPHB/L/d volumetric productivity, over fourfold higher than the best single-reactor configurations, with PHB up to 26% dcw. These results show continuous biomass availability, not maximal intracellular PHB content, is the decisive driver of volumetric productivity.Overall, this thesis highlights the value of environmental prospection and strain screening, since its microbiomes were assembled entirely from environmentally isolated strains. It demonstrates the need for strict control of culture conditions, as operational parameters, not population composition, governed PHB outcomes, and underscores the role of reactor configuration in translating biology into industrial productivity gains. Secondary effluent supporting biomass growth offers initial evidence that wastewater is a viable PHB feedstock, highlighting the potential of integrating wastewater treatment with biopolymer production.
  

DOCTORAL DEGREE IN GEOTECHNICAL ENGINEERING

• GUO, TIAN: Reducing the uncertainty on subsurface heterogeneity from microseismicity and geodetic data
  
  Author: GUO, TIAN
  Programme: DOCTORAL DEGREE IN GEOTECHNICAL ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 01/10/2026
  Reading date: 24/11/2026
  Reading time: 15:00
  Reading place: ETSECCPB.UPC, Campus NordBuilding C1. Classroom: 002C/Jordi Girona, 1-308034 Barcelona
  Thesis director: VILARRASA RIAÑO, VÍCTOR | WU, HAIQING
  Thesis abstract: Ensuring geological CO₂ storage (GCS) at scale requires more than just the availability of pore space. Thermal, hydraulic, mechanical, and chemical interactions during and after CO₂ injection pose challenges due to limited access to the subsurface. Reducing subsurface uncertainty is one of the most critical challenges for the safe deployment of GCS. This thesis aims to reduce this uncertainty by linking ground deformation and induced microseismicity with data-driven inversion, stress analysis, and numerical hydromechanical modeling. Specifically, this thesis evaluates (1) how ground deformation can be predicted and used to infer subsurface and fault properties, (2) how microseismic focal mechanisms can constrain the stress state and fault reactivation, and (3) how coupled simulators can support the long-term interpretation of the hydromechanical responses of Opalinus Clay to the injection of CO₂-rich brine.First, we develop a multi-stage, data-driven workflow to predict ground deformation caused by fluid injection into fault-intersected reservoirs. Synthetic poroelastic solutions are used for training and verified against CODE_BRIGHT simulations. Deep neural networks outperform linear regression and tree-ensemble-based models. Pressure change, shear modulus, and fault geometry are dominant controls on surface displacement, enabling rapid assessment of subsurface conditions. Secondly, we propose a probabilistic inversion framework based on conditional invertible neural networks (cINNs) to estimate subsurface parameters from geodetic data and quantify the non-uniqueness of the characterization. Synthetic tests demonstrate that the reservoir width is well-constrained and the throw-to-thickness ratio is partially constrained, whereas the fault dip is only weakly identifiable due to its subtle surface expression. Application of this framework to the Illinois Basin-Decatur Project (IBDP) reveals an equivalent reservoir width that increases over time, consistent with the lateral migration of the pressurized region.We also constrain the stress state and fault stability at the IBDP by integrating microseismic focal mechanisms with borehole data, vertical stress estimates, and instantaneous shut-in pressure (ISIP) measurements. Slip-tendency analysis of 26 focal mechanisms suggests a maximum horizontal stress (SHmax) magnitude of approximately 93 MPa, oriented at ~110°. Fault reactivation thresholds prove to be highly sensitive to ISIP, SHmax magnitude, and the fault friction coefficient.Finally, we present hydromechanical benchmark simulations based on the CO₂LPIE experiment at the Mont Terri Underground Rock Laboratory in Switzerland, comparing CODE_BRIGHT and OpenGeoSys in axisymmetric, anisotropic, and stratified configurations. Both codes reproduce consistent patterns of pressure diffusion and dissolved CO₂ transport, whereas the geomechanical fields (deformation, displacement, and stress) exhibit strong sensitivity to bedding anisotropy and constitutive models, highlighting the need for careful calibration against field data.Overall, this thesis demonstrates that GCS safety benefits from addressing monitoring, modeling, and uncertainty quantification in an integrated manner. Ground deformation provides spatial information on reservoir pressurization, microseismicity constrains stresses and fault reactivation, and machine learning models enable rapid prediction and probabilistic inversion for subsurface characterization. The proposed framework improves the characterization of subsurface structures and properties while supporting an adaptive assessment of GCS safety.
  

DOCTORAL DEGREE IN MATERIALS SCIENCE AND ENGINEERING

• FOOLADIMAHANI, SAGHAR: Assessment of fracture and fatigue behavior of cemented carbides by using small-scale artificial flaws
  
  Author: FOOLADIMAHANI, SAGHAR
  Programme: DOCTORAL DEGREE IN MATERIALS SCIENCE AND ENGINEERING
  Department: Department of Materials Science and Engineering (CEM)
  Mode: Article-based thesis
  Deposit date: 21/09/2026
  Reading date: 13/11/2026
  Reading time: 11:00
  Reading place: ESCOLA D'ENGINYERIA BARCELONA ESTC/Eduard Maristany, 16 (08019 Barcelona)EDIFICI A planta 0, SALA POLIVALENTmeet.google.com/swu-kiha-hwq
  Thesis director: LLANES PITARCH, LUIS MIGUEL
  Thesis abstract: This PhD thesis investigates the fracture and fatigue behavior of WC-Co cemented carbides through a deterministic damage-tolerance approach based on controlled artificial small-scale flaws. These materials, commonly referred to as hardmetals, are widely used in demanding engineering applications due to their combination of hardness, strength, and toughness. However, their structural reliability is strongly affected by microstructural heterogeneities and service-induced defects, which act as critical failure origins under monotonic and cyclic loading. Conventional assessment approaches rely mainly on stochastic natural flaws or long through-thickness cracks, thereby limiting reproducibility and predictive capability. This work addresses these challenges by developing and validating an experimental framework to systematically assess damage tolerance using well-defined artificial flaws.The research was conducted on four fine-grained WC-Co grades with different cobalt contents to isolate the influence of the metallic binder. Three complementary strategies were employed. Conical indentation reproduced blunt contact damage dominated by irreversible deformation and, at higher loads, radial cracking. Sharp Vickers indentation introduced surface fissures suitable for fatigue crack growth studies. Nanosecond laser ablation was optimized to generate reproducible microdimples and notch-like surface defects with controlled geometry. Their effects on residual strength, fatigue performance, crack growth behavior, and fracture toughness were assessed through mechanical testing, microscopy, fractography, crack-path inspection, and fracture-mechanics analysis.The results show that controlled artificial flaws provide a robust basis for evaluating damage tolerance in hardmetals. Under monotonic loading, residual strength is governed by flaw severity and intrinsic fracture toughness. For imprint-type or dimple-like defects without pre-existing cracks, strength reduction is mainly associated with geometric stress concentration. Once cracking occurs, fracture toughness becomes dominant, causing abrupt strength degradation in low-binder grades and more stable crack evolution in tougher materials. Under cyclic loading, the metallic binder plays a dual role. In crack-free damaged conditions, higher binder content increases fatigue sensitivity due to localized cyclic plasticity, whereas in pre-cracked conditions it improves resistance to subcritical crack propagation through crack-tip shielding and energy dissipation.A specific part of the thesis addresses laser ablation for fracture toughness determination. Laser-induced microdimples combined with quantitative fractography produced toughness values in close agreement with established fracture-mechanics methods. Laser-generated shallow notches also acted as controlled crack-initiation geometries for conventional toughness evaluation, providing a practical alternative to traditional pre-cracking procedures. These results demonstrate that laser-induced flaws are effective both for damage-tolerance studies and for reliable fracture-mechanics characterization.Overall, this thesis establishes a rigorous framework for investigating fracture and fatigue in WC-Co cemented carbides through controlled small-scale flaw engineering. The methodology distinguishes the effects of microstructure, damage condition, and flaw characteristics; clarifies the role of binder content under different loading conditions; and validates indentation- and laser-induced flaws for assessing strength, fatigue behavior, crack propagation, and fracture toughness. These findings provide a practical basis for reliability assessment and the design of hardmetals with improved resistance to service-like damage.
  

DOCTORAL DEGREE IN NETWORK ENGINEERING

• FERNÁNDEZ LANGA, SERGI: Future Immersive and Interactive Media Experiences: Multiparty Holoportation in Social Virtual Reality (VR) Scenarios
  
  Author: FERNÁNDEZ LANGA, SERGI
  Programme: DOCTORAL DEGREE IN NETWORK ENGINEERING
  Department: Department of Network Engineering (ENTEL)
  Mode: Normal
  Deposit date: 29/09/2026
  Reading date: 23/11/2026
  Reading time: 15:00
  Reading place: aula VGB107 de l'edifici Aulari (VGB) del Campus de Vilanova
  Thesis director: RINCON RIVERA, DAVID | MONTAGUD CLIMENT, MARIO
  Thesis abstract: The promise of Virtual Reality (VR) is that of a medium which makes you feel like being in a place where you are not and freely explore it, potentially immersed in a plot, as well as interact with any element in it. Beyond that, Social VR aims to provide togetherness, impacting users’ sense of presence, also through embodiment, and give the possibility to bring multiple people into the same story or context, into the VEs, so the experience is not only immersive but shared among a group of humans. This thesis aims primarily to design, develop, and evaluate new technological contributions for the application and scalability of multiparty holoportation in Social VR scenarios, with the general objective of improving the quality of shared immersive experiences. In other words, this work aims at evolving and improving state-of-the-art Social VR solutions, by enabling the integration of multiple remote users into shared VEs, embodied and represented as 3D volumetric holograms, captured and reconstructed in real-time. The thesis revolves around four main hypothesis: the first hypothesis is that VEs resulting from a strategic hybridization of synthetic 3D and 2D video-based elements can provide satisfactory levels of immersion, interactions and plausibility at low costs; the second hypothesis is that low-cost full-body volumetric capture setups can enable live multiuser holoportation in Social VR, while providing satisfactory levels of interaction, presence, embodiment and togetherness; the third hypothesis is that single camera volumetric capture setups can result in pleasant group-shared experiences (e.g., virtual meetings) in scenarios where the tele-ported users are not meant to move significantly; and the fourth hypothesis is that the bottlenecks that will arise when increasing the number of tele-ported volumetric users into a single VE can be mitigated by applying network-based media processing optimizations on the server side.The results from this thesis have resulted in three main research contributions: a full-fledged multiparty VR holoportation platform; a set of VR environments through which the developed technical components can be demonstrated and evaluated; and a server-side component to optimize multiuser holoportation or communications based on volumetric video.
  
• GONZALEZ DIAZ, NESTOR: Improving the Scalability and Availability of Indoor Location Solutions based on Wi-Fi RTT (IEEE 802.11mc)
  
  Author: GONZALEZ DIAZ, NESTOR
  Programme: DOCTORAL DEGREE IN NETWORK ENGINEERING
  Department: Department of Network Engineering (ENTEL)
  Mode: Normal
  Deposit date: 24/09/2026
  Reading date: 02/12/2026
  Reading time: 11:00
  Reading place: Sala de actos (001-Pg) de la EETAC (Castelldefels)
  Thesis director: ZOLA, ENRICA VALERIA | MARTÍN ESCALONA, ISRAEL
  Thesis abstract: Indoor positioning is a key enabler of location-based services, yet no technology has emerged as the indoor counterpart of Global Navigation Satellite Systems (GNSS). Wi-Fi Round-Trip Time (RTT), standardized in IEEE 802.11mc through the Fine Timing Measurement (FTM) protocol, is a promising candidate because it reuses the Wi-Fi infrastructure already deployed in most buildings and measures distances with sub-meter error on commodity devices.However, the real-world adoption of Wi-Fi RTT is hindered by three interrelated challenges. Accuracy degrades under Non-Line-of-Sight (NLoS) and multipath conditions, which introduce systematic biases in the ranging estimates. Scalability is capped by the peer-to-peer nature of the FTM protocol, which requires an exchange of location-dedicated frames between the user and the access point (AP). This procedure limits each AP to about thirty users updating their position once per second. Finally, the machine-learning models required to exploit the signal often exceed the memory and the latency budget of resource-constrained devices. This dissertation argues that accuracy, scalability, and computational complexity must be optimized jointly, meeting the memory, latency, and availability requirements of Internet of Things (IoT) deployments.The system proposed in this dissertation is based on fingerprinting and combines the active RTT observable with passive observables that add no network overhead, namely the Received Signal Strength Indicator (RSSI) and the standard deviation of the RTT burst. The system is organized into a clustering stage, a feature selection stage, and a multiclass classification stage. All experiments use measurement campaigns collected with commodity smartphones in two real scenarios: an open auditorium, with line of sight (Line-of-Sight, LoS) between users and APs, covered by up to seven 802.11mc-compatible APs, and a floor of corridors and offices, without line of sight (NLoS), served by three APs.The dissertation first characterizes the accuracy achievable with Wi-Fi RTT fingerprinting. Two RTT requests attain very good results in both scenarios: an accuracy above 97% in LoS when one RSSI is added, and a root-mean-square error (RMSE) between 50 and 65 cm in NLoS when three RSSI are added. It then proposes a clustering and feature selection architecture that distributes the FTM traffic overhead among the APs, increasing the potential number of concurrent users without degrading accuracy. Third, a low-complexity proximity model, which replaces the iterative clustering and the tree-based feature selection with a single sort of the observed RSSI, matches the accuracy of more complex algorithms while reducing the clustering memory to less than half and storing no persistent model; in addition, the RMSE stays below 30 cm in LoS and near half a meter in NLoS. Finally, it reveals that the RSSI from passive scanning and the RSSI from the FTM frames are systematically biased and locally uncorrelated observables that carry complementary positional information. The proposed system, evaluated with both sources, reaches an RMSE of 0.66 m at the highest user count and of 0.30 m with the three RTT requests in NLoS.Taken together, these results show that accuracy, scalability, and low computational complexity can be achieved jointly on commodity hardware, providing an integrated and reproducible framework for scalable and available indoor localization based on Wi-Fi RTT.
  
• MIRANDA PASCUAL, ÀLEX: Toward a Theory of Preprocessing and Composition in Differential Privacy
  
  Author: MIRANDA PASCUAL, ÀLEX
  Programme: DOCTORAL DEGREE IN NETWORK ENGINEERING
  Department: Department of Network Engineering (ENTEL)
  Mode: Change of supervisor
  Deposit date: 24/09/2026
  Reading date: 28/01/2027
  Reading time: 11:00
  Reading place: Karlsruhe
  Thesis director: PARRA ARNAU, JAVIER | STRUFE, THORSTEN
  Thesis abstract: Personal data processing is essential to many applications from health to finance and urban development, but data may contain sensitive information, raising serious privacy concerns. Privacy-enhancing technologies therefore seek to preserve utility while limiting disclosure, creating an inherent privacy–utility tradeoff. Differential privacy (DP) has become the standard privacy notion due to its strong protection guarantees and rigorous definition. Its composition properties allow complex DP mechanisms from simpler ones, with post- and preprocessing techniques extending DP to more tasks. However, DP can struggle to achieve an acceptable privacy–utility tradeoff for complex tasks. Moreover, the central DP model requires trust in a single curator, whereas local approaches remove this assumption at a substantial utility cost. Thus, this thesis studies how the main building blocks of DP—composition and preprocessing—affect the privacy–utility tradeoff and how this tradeoff adapts to the trust requirements. Understanding these tools helps researchers develop better DP mechanisms.We study the privacy-preserving publication of human trajectory databases under DP and syntactic notions. Our systematization shows that DP suffers under the increased complexity and presence of outliers. Because these limitations are not exclusive to trajectories, we focus the remainder of this thesis in general contexts.The negative impact of outliers motivates our study on record omission in DP as preprocessing. We first analyze sampling that amplifies privacy by uniformly selecting a subset of records of the database. However, our evaluation shows that core DP mechanisms with sampling consistently yield a worse tradeoff than the mechanism without sampling over all tested scenarios. We then introduce suppression in DP and derive theorems characterizing the privacy guarantees under arbitrary suppression, showing that it can either amplify or degrade privacy depending on the setting. However, targeted outlier suppression also does not improve the privacy–utility tradeoff and is less effective than sampling. These results show that preprocessing should be evaluated based on its overall privacy–utility impact, rather than privacy amplification alone.We next study composition and its interaction with preprocessing. Existing sequential and parallel composition theorems may overestimate privacy loss when mechanisms access overlapping data, while parallel composition does not apply to some DP variants, including bounded DP. We therefore introduce a general composition theorem that accounts for deterministic preprocessing under arbitrary domains and neighborhood definitions. By modeling data access as preprocessing, our theorems provide tighter privacy estimates and enable intermediate composition settings between sequential and parallel composition.Finally, we investigate limiting data access to reduce trust requirements. We introduce the k-curators model, positioned between the central and local models, where disjoint portions of the data are randomly distributed among k non-colluding curators. To analyze this model, we introduce probabilistic parallel composition, which provides tighter privacy bounds than standard parallel composition. Our evaluation of the privacy–utility–trust tradeoff shows that the k-curators model can limit the data accessed by each entity with minimal utility cost at fixed privacy levels.Overall, this thesis advances the understanding of DP by characterizing the effects of composition, preprocessing, and their interaction. It provides new theoretical results, more accurate privacy guarantees, and evaluates their impact on the privacy–utility(–trust) tradeoff, offering new insights and tools for designing future DP mechanisms.
  

DOCTORAL DEGREE IN PHOTONICS

• MINI DAS, APARNA: Investigation of Dynamic Catalyst -Electrolyte Interfaces in Electrochemical Systems for Establishing Activity Correlations.
  
  Author: MINI DAS, APARNA
  Programme: DOCTORAL DEGREE IN PHOTONICS
  Department: Institute of Photonic Sciences (ICFO)
  Mode: Normal
  Deposit date: 28/09/2026
  Reading date: 02/11/2026
  Reading time: 10:00
  Reading place: ICFO Auditorium
  Thesis director: GARCÍA DE ARQUER, FRANCISCO PELAYO
  Thesis abstract: Rising atmospheric CO2 levels, driven primarily by anthropogenic emissions, are a central contributor to global climate change and environmental instability. Electrocatalytic reactions provide a promising pathway for CO2 emission reduction through electricity-driven conversion of CO2 and/or water into value-added chemicals and fuels. In this context, catalysts are widely recognized as the key determinants of performance. This thesis takes the discussion a step further by examining how, in specific reactions, not only the catalyst itself but also the surrounding environment becomes crucial in defining performance matrices. More specifically, it investigates dynamic catalyst-electrolyte interfaces in water electrolysis and direct bicarbonate reduction, with a particular emphasis on interfacial dynamics and their correlation with activity.The thesis first outlines the fundamental aspects of water electrolysis and bicarbonate reduction, and then presents new concepts and findings on the role of dynamic catalyst-electrolyte interfaces in these systems. To do so, several in situ and operando techniques are employed, including Raman spectroscopy, X-ray absorption spectroscopy, electrochemical impedance spectroscopy, and quantitative bubble analysis, complemented by other electrochemical methods to establish correlations with performance.In the case of water electrolysis, the thesis focus on the oxygen evolution reaction, which represents the largest energetic penalty in the system. Correlations spanning atomic to microscale features are identified, collectively shaping the performance landscape of the reaction. Using complementary in situ probes, these correlations are shown to go beyond simple cause-and-effect relationships, highlighting the importance of uncovering the underlying mechanisms rather than relying solely on performance metrics.In direct bicarbonate reduction, performance is strongly influenced by the availability of CO2 at the interface, which is in turn linked to the local pH. Through operando Raman spectroscopy, a descriptor for the interfacial buffering state is identified, providing indirect insight into instantaneous CO2 availability. This hypothesis is further validated by engineering the catalyst-electrolyte interface to actively tune this descriptor, resulting in a threefold improvement in performance. This demonstrates how mechanistic understanding of the catalyst-electrolyte interface can guide informed hypotheses that are ultimately deterministic, reducing the experimental parameter space and improving optimization efficiency.Overall, the thesis emphasizes the importance of understanding why performance improves, rather than simply attributing progress to “better catalysts”. A deeper understanding of these systems, especially at high current densities, can lead to more coherent design principles and challenge existing assumptions about major factors governing the electrochemical performance. The results show that the catalyst surface cannot be considered a static entity, but rather a dynamic system whose structure and reactivity are continuously modulated by external factors. These insights provide a more detailed mechanistic picture of electrochemical interfaces and help identify key descriptors for catalyst design.
  
• PLASCENCIA OROZCO, SUSANA: Quantum memories based on rare-earth doped solids: from spatial multiplexing to new materials
  
  Author: PLASCENCIA OROZCO, SUSANA
  Programme: DOCTORAL DEGREE IN PHOTONICS
  Department: Institute of Photonic Sciences (ICFO)
  Mode: Normal
  Deposit date: 12/10/2026
  Reading date: 16/11/2026
  Reading time: 10:00
  Reading place: ICFO Auditorium
  Thesis director: DE RIEDMATTEN, HUGUES
  Thesis abstract: Quantum repeaters play a fundamental role in the development of long-distance quantum network. In most architectures, their functional core is constituted by quantum memories. Although many systems have been investigated as potential quantum memory, no single platform yet satisfies all the requirements simultaneously. Solid-state systems, particularly cryogenically cooled rare-earth doped crystals represent one of the most promising routes toward efficient, long-lived and massively multiplexed quantum memories, with great prospects for integration. This thesis addresses two main objectives. The first is to extend the spatial multimode capacity of a rare-earth-doped quantum memory using a praseodymium-doped yttrium orthosilicate Pr^{3+}:Y_2SiO_5 to increase the entanglement generation rate required for realistic quantum repeater application. The second objective, carried out in collaboration with Corning, investigates the spectroscopic properties of erbium-doped-chalcogenides to assess their suitability as fiber-integrated quantum memory operating at the telecom wavelength. To achieve spatial multiplexing in the Pr^{3+}:Y$_2SiO_5 crystal, we implement a quantum memory array (QMA) using a pair of acousto-optic deflectors, creating ten individually controllable spin-wave memory cell featuring on-demand read-out. By combing the spatial multiplexing of the array with the intrinsic temporal multimodality of the atomic frequency comb (AFC) protocol, we store and retrieve on-demand up to 250 modes using weak coherent pulses at the single-photon level, both surpassing the current state-of-the-art for solid-state quantum memory. The high signal-to-noise ratio and low cross-talks enable the storage of heralded single-photons. To do so, we use a cavity enhanced spontaneous parametric down conversion source to generate a pair of photons, with the signal photon stored in the QMA and the idler at telecommunication wavelength. We demonstrate the violation of the Cauchy-Schwarz inequality by at least 4.8 standard deviations in all the memory cells, demonstrating light-matter quantum correlations between the QMA and the telecom photon. We further demonstrate the potential use of our system as a random access quantum memory (RAQM) by storing path and time-bin qubits encoded with weak coherent states in the quantum memory array, achieving average fidelities of 95 ± 2% and 91 ± 2%, respectively. We further load sequentially a time-bin qubit in two different memory cells, simultaneously store both qubits, and perform a collective read-out. The high fidelities together with independent control pair represent a significant step toward a solid-state RAQM for quantum repeaters and photonic quantum processors. For the fiber-compatible quantum memory, we study three Er-doped chalcogenide glasses with a different glass network rigidities quantified by the coordination mean number (MCN). We measure their optical coherence time (T_2) through two-pulse photon echo (TPE) technique and find that the sample with MCN < 2.4 exhibits the shortest T_2, while the most rigid sample (MNC = 2.9) suffers from internal stress that also limits the coherence time. The intermediate rigidity sample (MNC = 2.69) shows a T_2 close to 1 μs (corresponding to a homegeneous linewidth ≈ 370 kHz) at 2 K and 0.2 T, which is among the longest coherence times reported for Er-doped glasses under practical cryogenic conditions, reachable with a standard close-cycle cryostat. However, the spectral hole burning measurements reveal hole widths around 50 MHz for two best TPE performing samples, severely restricting the achievable comb finesse and therefore constraining the AFC efficiency in these materials.
  
• PONS LANAU, ROGER: Single-Molecule Spatiotemporal Dynamics of Protein Export from the Endoplasmic Reticulum
  
  Author: PONS LANAU, ROGER
  Programme: DOCTORAL DEGREE IN PHOTONICS
  Department: Institute of Photonic Sciences (ICFO)
  Mode: Normal
  Deposit date: 06/10/2026
  Reading date: 12/11/2026
  Reading time: 15:00
  Reading place: ICFO Auditorium
  Thesis director: GARCÍA PARAJO, MARÍA | CAMPELO AUBARELL, FELIX
  Thesis abstract: The secretory pathway is a crucial process for cellular homeostasis, responsible for the synthesis, processing, and trafficking of approximately a third of all proteins to their final destinations. Newly synthesized secretory proteins enter the endoplasmic reticulum (ER) and, after proper folding, are sorted into and exported from specific ER subdomains named ER exit sites (ERES). As the first trafficking step, ER protein export plays a crucial role in ensuring the correct flow of proteins to downstream compartments, and defects result in a variety of pathological conditions. Unraveling its organization in space and time is essential to understanding its mechanisms. Despite the origins of this research field dating to the late 1960s, it was not until the widespread adoption of fluorescence microscopy techniques in the past three decades that rapid progress was made in how we conceptualize ER protein export in living cells. Yet, up until now, researchers have mostly relied on population-averaged measurements, which provide limited information on the molecular heterogeneity inherent in biology. Insights into molecular dynamics, transient interactions, short-lived states, or nanoscale distribution of secretory pathway proteins in living cells have largely remained veiled. Developing a complete view requires characterizing the process at the single-molecule level.In this thesis, we develop and apply a quantitative framework based on single-molecule fluorescence microscopy to dissect the dynamic organization of the early secretory pathway at the molecular level. By leveraging recent technological advances in fluorescent dyes, functional assays, gene editing, and machine learning tools, our approach enables the detection, monitoring, and analysis of individual cargo protein molecules as they are exported from ERES with high spatial precision (~20-30 nm) and temporal resolution (~10 ms). Our characterization provides novel insights into the orchestration of the early secretory pathway, allowing us to revisit conflicting ideas in the field. Overall, this thesis provides a generalizable methodology and an updated, more accurate view of a fundamental intracellular trafficking process.
  

DOCTORAL DEGREE IN POLYMERS AND BIOPOLYMERS

• MUÑOZ GALAN, HELENA: Sensor design and development for autonomous devices for disease diagnosis and therapy
  
  Author: MUÑOZ GALAN, HELENA
  Programme: DOCTORAL DEGREE IN POLYMERS AND BIOPOLYMERS
  Department: Department of Chemical Engineering (EQ)
  Mode: Normal
  Deposit date: 22/01/2026
  Reading date: pending
  Reading time: pending
  Reading place: pending
  Thesis director: ALEMAN LLANSO, CARLOS ENRIQUE | PÉREZ MADRIGAL, MARIA DEL MAR
  Thesis abstract: This doctoral research addresses key challenges in diabetes management by integrating sustainable materials, non-invasive sensing, and advanced insulin delivery technologies into a unified framework. Diabetes mellitus is a chronic metabolic disorder that requires continuous glucose monitoring and precise insulin administration to maintain glycemic control and reduce complications. Responding to these needs, the thesis is structured around three major contributions.First, the work advances the development of a previously created non-invasive glucose-monitoring device by incorporating recycled low-density polyethylene (LDPE) into the sensor design. The use of recycled LDPE improves sustainability, cost-efficiency, and environmental compatibility while maintaining reliable sensing performance. This demonstrates the feasibility of repurposing plastic waste in biomedical technologies without compromising functionality.Second, the research explores new strategies for controlled insulin delivery through stimuli-responsive hydrogels. Hydrogels based on poly(γ-glutamic acid) and multi-armed polyethylene glycol (PEG) were engineered to achieve sustained and tunable insulin release. The incorporation of poly(3,4-ethylenedioxythiophene) (PEDOT), a biocompatible conducting polymer, enables electrically triggered and on-demand insulin delivery. This approach offers a minimally invasive alternative to conventional injection-based therapies and highlights the potential of electro-responsive materials in smart drug-delivery systems.Finally, the thesis investigates the nanomechanical properties of multi-armed PEG hydrogels using a microcantilever-based optomechanical sensor. This analysis provides essential insights into the structural behavior, stability, and long-term performance of the hydrogels, contributing to the optimization of their mechanical robustness and responsiveness.Together, these contributions form a comprehensive strategy for improving diabetes care. By integrating non-invasive monitoring, environmentally conscious materials, and intelligent delivery platforms, the research promotes more sustainable biomedical solutions while advancing patient-centered therapeutic tools.
  

DOCTORAL DEGREE IN SIGNAL THEORY AND COMMUNICATIONS

• CHARI, SHREYA KRISHNAMA: Resource orchestration at the MEC for heterogeneous networks and services using Reinforcement learning enhanced by Generative AI
  
  Author: CHARI, SHREYA KRISHNAMA
  Programme: DOCTORAL DEGREE IN SIGNAL THEORY AND COMMUNICATIONS
  Department: Department of Signal Theory and Communications (TSC)
  Mode: Normal
  Deposit date: 01/10/2026
  Reading date: 03/12/2026
  Reading time: 09:30
  Reading place: A4-206, UPC campus Nord, Barcelona
  Thesis director: VERIKOUKIS, CHRISTOS
  Thesis abstract: The evolution of 5G and beyond communication has enabled rapid expansion of network capabilities, services and applications. The requirements of modern communication and computing infrastructures need to accommodate heterogeneous resource requests while maintaining service quality, resource distribution and energy efficiency. Shifting the compute-intensive applications to centralized cloud architectures has alleviated the problems related to network resource procurement. However, the geographical distance of these cloud data centers introduce communication delays, network congestion and privacy issues due to the physical distance from the source of the generated data. In light of these circumstances, edge computing emerged as a viable option to bring computation, storage, and intelligence closer to end users and other data sources. Multi-access Edge Computing (MEC) infrastructure makes it accessible for the communication networks to provide computing capabilities at or near the network edge. With resources available at the network edge, low-latency communication, context-aware applications and distributed learning is enabled. Despite the advantages, several significant issues such as task offloading, efficient resource allocation, service placement, and workload scheduling remain open for improvement. Most of the use cases expected in future wireless networks will present multi-objective and multi-constrained non-trivial optimization problems which require intelligent solutions. Moreover, the large-scale data generated at the user end with millions of connected devices and rapidly changing network environment requires a flexible solution that can adapt to the variability. This dissertation investigates smart solutions for MEC resource management while simultaneously accounting for changing network environment using artificial intelligence. The main aim is to develop algorithms that can address open challenges of orchestration and allocation of edge resources to dynamic network workloads. Further, the algorithms developed should also optimize the overall network performance by bringing down latency, energy and resource allocation gaps. Reinforcement learning (RL) is selected as the primary optimization method as it has the capability to interact and learn from custom environments that can mold to varying workloads and perform multi-objective optimization. Further, the RL feedback loop continuously provides learning signals to refine the decision-making strategy to improve resource management. On the whole, the contributions from this thesis encourage using RL methods for MEC resource orchestration and is further enhanced by inclusion of GenAI components. The findings establish that intelligent techniques can be used for dynamic coordination and allocation of heterogeneous edge resources for diverse services. The research work in this thesis, provides proof-of-concept resource management framework which is scalable, energy-efficient and adaptable to evolving requirements of future wireless networks.
  
• NAVARRO GESSÉ, EDGAR: Synthesis, Design and COM-based Modeling of SAW Electroacoustic Filters
  
  Author: NAVARRO GESSÉ, EDGAR
  Programme: DOCTORAL DEGREE IN SIGNAL THEORY AND COMMUNICATIONS
  Department: Department of Signal Theory and Communications (TSC)
  Mode: Normal
  Deposit date: 01/10/2026
  Reading date: 02/11/2026
  Reading time: 11:00
  Reading place: Aula C4-028-2, Edifici C4 de l'Escola d'Enginyeria de Telecomunicació i Aeroespacial de Castelldefels (EETAC) de la UPC
  Thesis director: COLLADO GOMEZ, JUAN CARLOS | MATEU MATEU, JORDI
  Thesis abstract: Radio-frequency filtering is a key function in modern wireless systems, where demanding requirements on selectivity, bandwidth, compactness, and integration must be satisfied within a limited physical footprint. Surface acoustic wave (SAW) devices provide an attractive solution because they combine strong frequency selectivity with compact implementation. However, as filtering functions become more complex and conventional ladder architectures are complemented by multimode and hybrid configurations, the design problem extends beyond the synthesis of an equivalent electrical network: synthesized resonances, couplings, capacitances, and impedance levels must ultimately be translated into a physically realizable acoustic structure subject to the electroacoustic coupling, dispersive behaviour, geometrical constraints, and fabrication limitations of the selected technology.This thesis addresses the absence of a sufficiently systematic bridge between a prescribed filtering function and the final geometry of advanced SAW filters. The central approach is a synthesis–technology-adaptation–Coupling-of-Modes (COM) framework in which network synthesis establishes the required resonant and coupling conditions, topology-dependent circuit transformations or resynthesis enforce physical compatibility, and the resulting targets are mapped onto distributed COM structures that account for the physical dimensions of the interdigital acoustic geometry. The objective is not to replace physical modelling with an ideal equivalent circuit, but to preserve an explicit and repeatable correspondence between the prescribed response, a technology-compatible circuit representation, and the final acoustic geometry.At synthesis level, a unified driving-point extraction procedure combines finite-transmission-zero nonresonating-node sections with transmission-zero-at-infinity resonating-node sections, allowing a single filtering function to be partitioned between independent ladder resonators and multimode coupled-resonator-filter (CRF) sections. An enforced-zero deflation technique improves the numerical robustness of high-order synthesis by restoring the exact prescribed immittance root before each polynomial division. For conventional ladder filters, the Butterworth–Van Dyke-to-COM mapping is extended to a synthesis–mapping–resynthesis loop that accounts for the dispersive electroacoustic coupling of the technology.For multimode structures, a circuital Nakamura COM formulation and inverse-design procedures are developed for isolated two- and three-mode CRFs and extended to mixed ladder–CRF and cascaded-CRF architectures. The resulting design flow combines topology-specific circuit transformations with COM-based inverse design, incorporating the required impedance scalings, inverter transformations, static-capacitance compensation, technology-aware resynthesis, and physical partitioning needed to obtain realizable structures and ultimately determine their geometric dimensions. The proposed synthesis and circuit-transformation procedures are formulated and implemented as systematic and repeatable numerical procedures, while CRF inverse design and COM evaluation are implemented in a dedicated MATLAB design tool. The methodology is validated through COM simulations and measurements of fabricated ladder and two-mode CRF devices. A complementary tunable in-line filter further demonstrates the broader applicability of synthesis-based electroacoustic filter design, with measured continuous centre-frequency tuning from 0.99 to 1.03 GHz and up to 100% bandwidth control.Overall, the thesis establishes a systematic but topology-dependent methodology for translating prescribed filtering functions into physically realizable electroacoustic filter geometries, extending synthesis-driven design from conventional SAW ladders to multimode, mixed, cascaded, and reconfigurable architectures.
  

DOCTORAL DEGREE IN SUSTAINABILITY

• ZHANG JIN, TIANYU: From policy ambition to organisational practice. Understanding food waste interventions effectiveness in healthcare, hospitality and retail.
  
  Author: ZHANG JIN, TIANYU
  Programme: DOCTORAL DEGREE IN SUSTAINABILITY
  Department: University Research Institute for Sustainability Science and Technology (IS.UPC)
  Mode: Normal
  Deposit date: 01/10/2026
  Reading date: 09/11/2026
  Reading time: 12:00
  Reading place: SALA DE GRAUS Campus del Baix Llobregat, Edifici D4Carrer Esteve Terradas, 8, 08860 Castelldefels (Barcelona).
  Thesis director: GIL ROIG, JOSE MARIA | RAHMANI MEDDOUR, DJAMEL
  Thesis abstract: Food waste is one of the main challenges to the sustainability of the agri-food system. However, the lack of high-resolution data and context-specific evidence limits the effectiveness of policy and management responses. This thesis approaches food waste as a wicked problem: one that is interconnected across multiple scales, resistant to universal solutions, and shaped by the organisational contexts in which it is generated. Its central argument is that these conditions determine both which interventions can be implemented and the magnitude and distribution of the changes observed.The thesis examines interventions across three contrasting organisational contexts: a tertiary hospital, two luxury hotels, and a regional supermarket chain. Drawing on Scott’s (2008) institutional framework, sector-specific literature, and empirical observations, it uses abductive analysis to develop a provisional, cross-sectoral typology of organisational constraint architectures.Each case is based on high-resolution primary data and mixed methods. The hospital study combines comprehensive direct weighing across the entire food service system during two one-week measurement campaigns, interviews, and a causal loop diagram to evaluate a digital dietary management system and an awareness-raising intervention. The hotel study combines pre- and post-intervention measurements, qualitative assessments, and sensor-based monitoring in two establishments to evaluate operational kitchen interventions. The retail study analyses 219,032 product-level food waste records collected over fifteen months, using Wilcoxon tests, two-way fixed-effects panel regression, and two co-created consumer-facing interventions.In the hospital, the intervention period was associated with an extrapolated annual difference of 48,668 kg (11.9%) and approximately 187 tonnes of CO₂ equivalent, based on the study assumptions. However, attribution is limited by the uncontrolled pre-post design, the short duration of the measurement campaigns, the change in sampling strategy, and the simultaneous implementation of the interventions. In the hotels, food waste generated during kitchen preparation decreased by up to 81%, whereas customer-facing strategies were rejected because they were considered incompatible with the existing service model. Changes in plate waste demonstrate the need to assess both prevention and the potential displacement of waste at the system level. In the retail case, the consumer-facing interventions produced limited aggregate changes. Overall, 52.9% of losses were concentrated within the first two days after delivery, a pattern consistent with initial inventory imbalances and limited short-term exposure to consumer demand, although other operational explanations cannot be excluded.The thesis makes four contributions. First, it provides a comparative body of granular empirical evidence on food waste across the hospital, luxury hotel, and retail contexts studied in Spain. Second, it develops a provisional typology of constraint architectures that can be used to assess intervention feasibility before implementation. Third, it documents a structured co-creation framework that can be adapted to design context-sensitive interventions. Finally, it reframes food waste reduction as a governance problem: the relevant question is not only whether interventions work in principle, but whether they can be deployed within the organisational conditions of each context.
  

DOCTORAL DEGREE IN THEORY AND HISTORY OF ARCHITECTURE

• ALAZETTA, PASCALE ISABELLE: Quartiers anciens en déshérence - pratiques et expérimentations du neutre
  
  Author: ALAZETTA, PASCALE ISABELLE
  Programme: DOCTORAL DEGREE IN THEORY AND HISTORY OF ARCHITECTURE
  Department: Department of History and Theory of Architecture and Communication Techniques (THATC)
  Mode: Change of supervisor
  Deposit date: 29/09/2026
  Reading date: 20/11/2026
  Reading time: 09:00
  Reading place: Sala de Actos del Institut d’Urbanisme et de Géographie Alpine. Grenoble.
  Thesis director: ROSSELLO NICOLAU, MARÍA ISABEL | SANTAMARIA, FREDERIC SEBASTIEN | SCARNATO, ALESSANDRO
  Thesis abstract: Along the French Mediterranean coast, the central districts of medium-sized cities have, in recent years, seen their populations change, the mix of social classes disappear, the businesses close. They have sometimes become places of social marginalization where unsanitary conditions take hold. Attempts are being tried to reverse this trend. These attempts are often adequate in the central, vibrant parts of cities, where the refurbishment of public spaces and facilities, combined with the restoration of building facades and the renovation of historic buildings, revitalizes these historic centers and generates a certain commercial, cultural, and social vitality. The challenges lie in older working-class neighborhoods, where services and businesses rarely establish themselves, where public spaces receive little improvement, and where modest housing is difficult to adapt to contemporary living standards due to high building densities with cramped units, lack of light, ventilation, private outdoor space, accessibility, greenery, parking spaces, etc. Yet they consist of urban forms and historic buildings that possess significant spatial, aesthetic, and social qualities, as well as adaptability to topography, climate, and construction. Often, the solutions proposed trivialize the space, as interventions are strictly functional and overlook sensitive aspects of urban composition and building design. Contemporary architecture can also be controversial in these areas due to heritage constraints. And when the revitalization of these neighborhoods is successful, it often leads to their gentrification. It is therefore necessary to consider how these complexes might be adapted to contemporary lifestyles and ways of living, in an era where environmental concerns must necessarily guide solutions. The question of public tools and policies aimed at run-down historic neighborhoods—their content, implementation, complementarity, interconnection, and the issue of governance—must also be examined.This project proposes to analyze some of these situations and seek solutions that would allow for intervention and find a comfort that meet the expectations of contemporary life for a diverse population, while countering the mechanisms of gentrification. This work will be based on the specific case of four districts in medium-sized French cities with a rich urban and architectural heritage—Béziers, Narbonne, and Perpignan—which have been increasingly confronted in recent years with a marked decline in the quality of life in parts of their historic centers. Despite growing signs of poverty—which exceed national averages—these cities, unlike other struggling cities and due to their location along the Mediterranean coast, are experiencing population growth, rising real estate prices, and strong tourist appeal.In contrast, the city of Girona, in Spanish Catalonia, which has been the subject since the 1980s of an ambitious renovation plan that has led to an undeniable revitalization of its historic center, appears to be facing problems linked to gentrification and a proliferation of tourist accommodations that have altered the lives of its residents. In response to overly procedural specifications, programs, and spatial solutions, a work on neutral practices in urban planning and architecture may be explored. The neutral, as Roland Barthes understands it, can be understood here as the desire to leave spaces open, resistant to theoretical interpretations and subjective intentions that predetermine meaning and freeze the possibilities of inhabiting.
  

DOCTORAL DEGREE IN APPLIED MATHEMATICS

• MARTIN LLOPIS, ALVARO: Discrete Schrödinger Operators for Random Walks on Networks.
  
  Author: MARTIN LLOPIS, ALVARO
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN APPLIED MATHEMATICS
  Department: School of Mathematics and Statistics (FME)
  Mode: Normal
  Deposit date: 27/10/2026
  Deposit END date: 09/11/2026
  Thesis director: CARMONA MEJIAS, ANGELES | JIMÉNEZ JIMÉNEZ, MARIA JOSE
  Thesis abstract: This thesis explores the connections between spectral graph theory, discrete potential theory, and stochastic processes, using the discrete Schrödinger operator, in its matrix form, to analyze random walks on networks. Classical approaches to network dynamics often rely on the standard combinatorial or probabilistic Laplacian. In contrast, this work generalizes these methods by introducing a discrete Schrödinger matrix, Lq, which incorporates a potential vector q. This extension provides a flexible and robust framework for modeling random walks on non-homogeneous networks, allowing for the inclusion of intrinsic node weights, absorption costs, or local penalties.From an analytical perspective, the core methodology relies heavily on the use of generalized inverses; specifically, on the orthogonal Green matrix, which is nothing more than the group inverse of the Laplacian matrix. We demonstrate that the group inverse contains information about both the structure and the dynamics of the network.The main contributions of this research are the derivation of matrix expressions for two fundamental network parameters, the Mean First Passage Time (MFPT) and Kemeny’s constant. We present formulas obtained from the properties of the associated operators.Ultimately, this thesis provides analytical tools to compute global transport parameters, opening new avenues for understanding diffusion, navigability, and vulnerability in networks.
  

DOCTORAL DEGREE IN ARTIFICIAL INTELLIGENCE

• GRAU ESCOLANO, JORDI: Bikes and Bytes: Data-Driven Methods for Planning and Operating Docked Bike-Sharing Systems
  
  Author: GRAU ESCOLANO, JORDI
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN ARTIFICIAL INTELLIGENCE
  Department: Department of Computer Science (CS)
  Mode: Normal
  Deposit date: 27/10/2026
  Deposit END date: 09/11/2026
  Thesis director: VICENS BENNASAR, JULIAN ANTONIO
  Thesis abstract: Docked bike-sharing systems (BSSs) are an accessible, low-carbon mobility option that can replace short urban car trips and improve connectivity to public transport. As BSSs have evolved into large-scale services, performance increasingly depends on decisions across the system life cycle, from station placement to day-to-day operations. This thesis develops data-driven methods to support these tasks, using Barcelona and its docked BSS, Bicing, as case studies.First, the thesis addresses station network design. Existing optimization approaches often fail to quantify system accessibility realistically, transparently explore trade-offs among diverse planning objectives (e.g., maximizing demand coverage and social equity), or explicitly control station layout, which is typically a by-product of site selection. This thesis presents a flexible planning framework combining modular multi-criteria scoring with directed street-network accessibility metrics adjusted for topography, and uses a genetic algorithm to optimize station layouts while explicitly controlling inter-station proximity and overall accessibility. Applied to Barcelona, it identifies near-optimal configurations for diverse scenarios and shows that altitude-adjusted accessibility helps correct service gaps in hilly districts. The framework is also used to identify new stations that complement an existing network. Turning to day-to-day operations, the thesis first characterizes Bicing’s mobility patterns, which underpin two predictive use cases. The first one targets maximizing electric bike availability by keeping batteries charged without exact route information or in-transit battery measurements. The thesis proposes a joint forecasting framework coupling a Markov-chain mobility model with a physics-based battery model to simulate trajectories and energy states, yielding accurate battery estimates and reliable short-horizon station-level forecasts that enable proactive charging strategies. The second use case also targets increasing availability, but by reducing downtime through anticipation of bike component failures. The thesis develops predictive maintenance models for three key components (brake pads, wheel spokes, and chains) based on survival analysis and machine learning, using maintenance records, historical mobility indicators, and weather variables to estimate failure risk. The models achieve consistently strong predictive performance, and interpretability analysis shows that usage intensity is the main driver of degradation, with higher average speeds and electric propulsion further shortening component lifetimes. Overall, the thesis shows that docked BSS planning and operations can be improved with computational, data-driven decision support even under data limitations. The proposed methods enable a shift from heuristic planning and reactive management to evidence-based, anticipatory strategies that are transferable to other cities with comparable spatial and operational data.
  

DOCTORAL DEGREE IN CIVIL ENGINEERING

• BOPP IBAR, MARIA JESUS: Adopción de tecnologías de videojuegos y realidad extendida (XR) en la enseñanza de ingeniería civil
  
  Author: BOPP IBAR, MARIA JESUS
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN CIVIL ENGINEERING
  Department: Barcelona School of Civil Engineering (ETSECCPB)
  Mode: Normal
  Deposit date: 28/10/2026
  Deposit END date: 10/11/2026
  Thesis director: MORA SERRANO, FRANCISCO JAVIER | VALERO GARCIA, MIGUEL
  Thesis abstract: This doctoral thesis addresses the gap between the rapid technological evolution of industry, driven by the Construction 4.0 paradigm, and the persistence of traditional teaching methodologies in civil engineering education. In this context, the limited adoption of interactive technologies in education is identified as the central problem, a situation attributable not only to technological barriers but primarily to human and institutional factors, particularly lecturers’ lack of time and technical training to design and implement these experiences.To address this problem, the research is structured into three phases. The first consists of a diagnostic assessment conducted using a mixed-methods approach, combining semi-structured interviews with 20 university lecturers, mainly from Spain and Chile, with surveys administered to lecturers and students in both countries. The interviews identified a lack of time and technical knowledge as the main barrier to incorporating interactive technologies into teaching. The quantitative data obtained through the surveys were analysed using an adaptation of the UTAUT2 model, aimed at identifying the factors influencing lecturers’ and students’ intention to adopt video game–related technologies. The results show that performance expectancy is the main factor associated with this intention to use.Following the diagnostic assessment, the development of The Death Neglect, a virtual reality experience aimed at occupational risk prevention on construction sites, is presented. Its evaluation was conducted at two technology fairs. Its design and development process helped identify challenges and lessons learned that contributed to the formulation of a methodological proposal.Next, the RGB framework (Roadmap, Gameplay Construction, and Beta Testing) is presented, a methodology for designing serious games that integrates instructional design principles with video game design methodologies. Its construction and refinement drew on the experience gained during the development of The Death Neglect and MohrPlay through an iterative process.Subsequently, the development of MohrPlay, a mobile video game intended to teach Mohr’s circle, is presented. It was evaluated on two occasions at the Pontificia Universidad Católica de Valparaíso. This case illustrates the design decisions and lessons learned that underpin the RGB methodology.The research culminates in the RGB-A proposal, an extension of RGB that organises the foundations of instructional design and video game design through a dual conceptual hierarchy of five levels. Its application is illustrated through the review and reformulation of MohrPlay mechanics, whose implementation and evaluation are proposed as future work.
  

DOCTORAL DEGREE IN COMPUTATIONAL AND APPLIED PHYSICS

• MARCH PONS, DAVID: Collective Decision-Making in Bio-Inspired Swarms: From Behavioral Rules to Interaction Structure
  
  Author: MARCH PONS, DAVID
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN COMPUTATIONAL AND APPLIED PHYSICS
  Department: Department of Physics (FIS)
  Mode: Normal
  Deposit date: 26/10/2026
  Deposit END date: 06/11/2026
  Thesis director: PASTOR SATORRAS, ROMUALDO | MIGUEL LOPEZ, M. DEL CARMEN
  Thesis abstract: Living groups, from human societies to insect colonies, self-organize and coordinate in order to thrive in nature. Collective decisions are a central task in their life cycles, keeping the group cohesive around social norms, foraging strategies or collective movement. This self-organization emerges from the individual behaviors and social interactions between members of the group, so identifying the mechanisms that let a group gather and exchange information and reach a decision outcome is a central question in the study of collective behavior.The self-organizing capacity of insect societies has, in turn, long inspired the design of decentralized protocols for swarm robotics: combining insights from nature with engineering principles allows artificial swarms to reach collective decisions---on tasks such as effort allocation or objective selection---without any external control. Studying how such collective behaviors emerge from the interactions between the units of a group sits at the core of complex systems science, and it is within this framework that the present thesis addresses collective decision-making.Specifically, we investigate mechanisms of collective decision-making inspired by the social behavior of honeybee swarms endowed with a nest site selection task: a paradigmatic example in which individual behavior---information seeking, option quality assessment---and social interactions---recruitment of uncommitted peers, inhibition of divergent opinions---intertwine to produce consensus as an emergent property of the swarm. From these traits we extract basic modelling assumptions, studying the resulting agent-based models through both mean-field analytical methods and stochastic simulations.We first examine how the balance between individual and social information-seeking, and between competing option qualities, shapes decision accuracy and speed. We further show that the best outcomes are attained at the onset of criticality, and that the model replicates superorganism-like features, with the swarm reacting to physical stimuli as a single cognitive unit. Then, we implement this model on a swarm of Kilobot robots, complemented by virtual simulations, to establish how the percolation of a proximity-based, time-varying communication network---governed by interaction duration, range and robot density---shapes the emergence of consensus under real sensing and communication constraints. Building on this, we introduce a non-linear cross-inhibition signal that captures a many-body effect, whereby an agent abandons a divergent opinion only after receiving a sufficient number of inhibitory signals. We show that this signal enhances consensus accuracy and decision speed, at the cost of reliably selecting the higher-quality option. Finally, we address the complementary symmetry-breaking problem, by which the group must resolve a decision deadlock between equivalent options. We show that higher-order group interactions---whose effect is not reducible to a sum of pairwise ties---offer an alternative to cross-inhibition, giving rise to a qualitatively different landscape of first- and second-order phase transitions.Together, these results delineate how individual behavioral rules and the structure of social interactions---time-varying, pairwise or higher-order, linear or non-linear---shape the speed, accuracy and robustness of collective consensus. Beyond honeybee swarms, this work contributes theoretical tools and design principles for decentralized decision-making in artificial swarms, bridging mean-field theory, network science and embodied robotic experimentation.
  
• MONTALÀ SALES, RICARD: Data-Driven Strategies for Active Flow Control in Wings
  
  Author: MONTALÀ SALES, RICARD
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN COMPUTATIONAL AND APPLIED PHYSICS
  Department: Department of Physics (FIS)
  Mode: Article-based thesis
  Deposit date: 21/10/2026
  Deposit END date: 03/11/2026
  Thesis director: SORIA GUERRERO, MANUEL | RODRIGUEZ PEREZ, IVETTE MARIA
  Thesis abstract: Flow separation remains a major limitation to wing aerodynamic performance, particularly at low speeds, where viscous effects become increasingly important and reduce the ability of boundary layers to withstand adverse pressure gradients. Active flow control (AFC) offers the possibility of mitigating these effects by modifying the flow through localized actuation. However, designing effective AFC strategies becomes increasingly challenging for complex separated flows, for which the relationship between actuation, flow dynamics, and aerodynamic performance is highly nonlinear. This doctoral dissertation investigates data-driven methodologies, combined with high-fidelity computational fluid dynamics (CFD), for the autonomous discovery of AFC strategies in increasingly complex aerodynamic configurations.The research first establishes a detailed physical understanding of the turbulent flow past the three-element 30P30N high-lift wing using wall-resolved large eddy simulations (LES) at a Reynolds number of Re=750,000. The analysis identifies the main flow mechanisms governing the aerodynamic performance of the configuration, including boundary-layer development, separated shear layers, wake dynamics, and the interactions between the slat, main element, and flap. Conventional predetermined AFC strategies are subsequently investigated, demonstrating their ability to modify the flow while highlighting the difficulty of identifying effective actuation parameters solely from physical intuition.To overcome these limitations, a computational framework coupling high-fidelity CFD with deep reinforcement learning (DRL) is developed and systematically validated across increasingly complex flow configurations. The framework is first validated for the canonical flow past a circular cylinder at Re=100, yielding results consistent with those reported in the literature. It is then applied to the NACA0012 airfoil at Re=1,000 and high angle of attack, where the flow exhibits weakly turbulent characteristics. An extensive analysis of the control strategy learned by the DRL agent demonstrates its ability to recover actuation mechanisms consistent with established flow-control strategies reported in the literature. The methodology is subsequently applied to the fully separated SD7003 airfoil at Re=60,000, where the learned control policies considerably outperform predetermined AFC strategies. Furthermore, wall-mounted sensors are shown to achieve comparable control performance to wake-based sensing while requiring fewer sensors, providing a more realistic sensing configuration for practical AFC applications.Finally, the methodology is applied to the 30P30N high-lift wing initially studied, revealing the computational challenges associated with DRL at higher Reynolds numbers. The reduced time step required by the high-fidelity CFD simulations, combined with the large number of flow-control experiences required for DRL training, results in substantial computational costs and complicates the selection of suitable control configurations and hyperparameters. Bayesian optimization (BO) is therefore investigated as an alternative data-driven methodology for the low-dimensional control problem. After only 19 CFD evaluations, BO identifies an effective actuation strategy that maintains the baseline lift while reducing drag and increasing aerodynamic efficiency. Compared with the corresponding DRL training campaign, BO requires approximately half the wall-clock time and nearly sixty times fewer GPU-hours.
  

DOCTORAL DEGREE IN COMPUTER ARCHITECTURE

• MORAIS, LUCAS HENRIQUE: Hardware-Accelerated Task Management for Fast Dependence and Locality-Aware Task Scheduling
  
  Author: MORAIS, LUCAS HENRIQUE
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN COMPUTER ARCHITECTURE
  Department: Department of Computer Architecture (DAC)
  Mode: Normal
  Deposit date: 27/10/2026
  Deposit END date: 09/11/2026
  Thesis director: ALVAREZ MARTINEZ, CARLOS | JIMENEZ GONZALEZ, DANIEL
  Thesis abstract: High-performance computing systems have become increasingly parallel because single-core performance scaling no longer provides enough throughput by itself. Exploiting those systems is difficult when applications are written as mostly sequential programs: the runtime must uncover the parallel work that the programmer exposes and keep many cores busy without spending more time scheduling than computing. Task parallelism addresses that problem by representing independent or dependency-constrained pieces of work explicitly, and fine-grained tasking can expose even more parallelism if its management overhead is kept low.This thesis follows a single scalability problem through three hardware design steps. Fine-grained Task Scheduling can expose parallelism by dynamically assigning ready tasks to cores, but the scheduling machinery itself must be fast enough not to consume the parallelism it reveals. The first step therefore demonstrates low-latency hardware support for dynamic task scheduling on a large RISC-V multi-core system, establishing the operating regime in which hardware assistance can make fine-grained scheduling practical.Fast scheduling, however, does not by itself preserve data locality: once a task has all its dependencies fulfilled, executing it on the first available core can separate it from the tasks that produce the data it depends on. The second step addresses this locality cost by using hardware-accelerated Label Propagation (LP) to cluster tasks with intersecting data patterns and guide their placement on the same core. Applied to synthetic benchmarks and real-world workloads running on a 24-core RISC-V processor mapped to the Alveo U55C FPGA, this approach improves overall program execution times by up to 1.50x and reduces average measured task execution cycles by up to 1.81x. These gains rely on the latency-oriented nature of the accelerator, which will typically cluster dynamic task graphs in under 300 cycles and, in processor-cycle terms, up to 581x fewer cycles than the compared software implementation on its x86 platform. By treating Task-LP's output as locality hints rather than hard constraints, the system improves locality on evaluated memory-sensitive workloads, while relaxation allows non-preferred placement when strict locality would otherwise leave cores idle.Having shown that LP can improve task placement only when clustering is fast enough for the scheduler, the thesis then asks what changes when the limiting concern becomes graph capacity. The third step, *BlazeCount*, is an FPGA-implemented, capacity-oriented hardware design point for graphs whose stored edge list and labels fit in UltraRAM. It keeps those structures on chip, streams edges through a 512-bit AXI interface, dispatches neighbor reads through conflict-aware label-memory logic, and computes modal labels (the most frequent neighbor labels) with a bounded two-tier approximate counter. BlazeCount is a stand-alone Alveo U250 architectural prototype driven by a minimal XRT C++ host harness. Its role is to define the capacity side of the same LP hardware design space. The resulting design point informs future scheduler-integrated LP engines sized to balance latency and graph capacity, while accounting for the diminishing returns to parallelism from ever larger task graphs.
  

DOCTORAL DEGREE IN CONSTRUCTION ENGINEERING

• CARRASCO ASTUDILLO, NICOLÁS: Adoption of the water-energy-food nexus to improve buildings’ climate resilience
  
  Author: CARRASCO ASTUDILLO, NICOLÁS
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN CONSTRUCTION ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 29/10/2026
  Deposit END date: 11/11/2026
  Thesis director: GASPAR FÀBREGAS, KÀTIA | CASALS CASANOVA, MIQUEL
  Thesis abstract: Rising temperatures are increasingly affecting people's health, water consumption and food production, with direct consequences for the built environment. As people spend most of their time indoors, buildings are particularly exposed to the impacts of climate change, including hazardous events such as overheating, compromising occupant comfort, increasing the cooling energy demand, and posing direct health risks. To address these challenges, the concept of resilience has been developed, leading to a range of retrofit solutions that have achieved satisfactory results in reducing overheating, lowering emissions and energy demands, and improving the management of critical resources. However, these solutions tend to address challenges in an isolated way, while climate change demands more holistic responses.One concept that could address this challenge is the water-energy-food (WEF) nexus. By encouraging the synergy between these critical resources, populations experiencing water scarcity or food insecurity have been able to manage them more efficiently. Although there have been advances in implementing this concept in the built environment, no systematic method exists for evaluating WEF integration at the building scale from the point of view of resilience.Thus, this doctoral thesis adapts the water-energy-food nexus approach to the context of improving the climate resilience of buildings, assessing the applicability and effectiveness of the concept. Through a review of resilience and WEF metrics, a quantitative framework was developed to evaluate retrofit solutions in terms of both resilience improvement and resource self-sufficiency, using the water-energy-food nexus index for buildings. To test this framework, different rooftop retrofit strategies were applied to reference residential buildings in Catalonia, Spain, modelled in DesignBuilder and simulated under different climatic conditions in the EnergyPlus engine.The main results show that traditional solutions address overheating and energy efficiency effectively, but that resource self-sufficiency is largely absent from their design logic. The exception is the combination of an agrivoltaic system with rainwater collection, which improves resilience indicators while achieving WEF integration consistently across climate zones and future scenarios. This thesis demonstrates that climate resilience and resource self-sufficiency are not conflicting objectives but complementary ones, and provides researchers, policymakers and decision-makers with a method and empirical evidence to support their joint evaluation in building retrofit decisions.
  

DOCTORAL DEGREE IN ELECTRIC ENERGY SYSTEMS

• HENAO MUÑOZ, ANDRES CAMILO: Modeling, Control and Stability Analysis of Grid-Connected Multiport Converters
  
  Author: HENAO MUÑOZ, ANDRES CAMILO
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN ELECTRIC ENERGY SYSTEMS
  Department: Department of Electrical Engineering (DEE)
  Mode: Normal
  Deposit date: 27/10/2026
  Deposit END date: 09/11/2026
  Thesis director: LUNA ALLOZA, ALVARO | PEPICIELLO, ANTONIO
  Thesis abstract: The increasing penetration of power-electronic-interfaced generation, energy storage, and flexible loads is progressively shifting power-system dynamics toward converter-control-dominated operation. Multiport power converters offer a compact and efficient solution for integrating multiple energy resources and interconnecting AC and DC networks. However, the electrical coupling among their ports may give rise to control interactions that affect regulation and stability. Despite their potential benefits, research on multiport converters is still at a relatively early stage, and their modeling, control, and stability in grid-connected applications require further investigation. Motivated by this, this thesis investigates different modeling and control approaches for multiport converters, grid-connected converters, and grid-connected multiport converters. Moreover, a modeling framework to develop models suitable for small-signal stability analysis is developed.The isolated triple active bridge (TAB) converter is investigated as the main multiport topology. Existing modeling approaches show that active-power estimation accuracy is considerably sensitive to the operating point and model formulation. Moreover, an Improved First-Harmonic Approximation generalized average model is proposed, providing more accurate power-transfer estimation and reduced operating-point sensitivity, preserving model order and computational complexity. Moreover, the inherent power coupling among TAB ports is addressed through model-based linear decoupling and Linear Active Disturbance Rejection Control (LADRC). A comprehensive comparison shows that LADRC provides stronger rejection of cross-coupling effects and greater robustness to modeling uncertainties than conventional matrix-decoupling approaches.To interface the TAB converter and the AC grid, two-level converters are considered. Both grid-following (GFL) and grid-forming (GFM) control strategies are investigated, with particular emphasis on power-synchronization control loops (PSL) for GFM converters. The damping limitations of conventional swing-equation-based virtual synchronous generator (VSG) control are analyzed, and relationships and complementarities between VSG-based PSL and alternative phase-locked-loop (PLL), PI-, and IP-based PSL are established. Based on these relationships, two PLL-free PSLs, denoted PI-VSG and IP-VSG, are proposed. Analytical demonstrations, simulations, and experimental results validate enhanced design flexibility, transient response, and synchronization robustness during power-reference changes, grid-frequency deviations, voltage sags, and variations in grid strength.Finally, a modular small-signal modeling methodology based on the Component Connection Method is developed for the TAB-based soft open point. Standalone converter, control, and network models are developed and hierarchically interconnected to represent different GFL and GFM configurations while preserving the dynamic relationships between the AC and DC sides. The validated models are used for eigenvalue, participation-factor, and parametric sensitivity analyses. The results show that SOP stability depends jointly on converter control, AC network impedance, power-flow magnitude and direction, and AC-DC coupling. In GFL-GFL operation, the critical instability mechanisms are primarily associated with mid-frequency interactions and are particularly sensitive to the grid conditions when the grid-connected converter control mode includes a DC voltage controller. Introducing GFM control increases robustness against these interactions and shifts the critical dynamics toward lower-frequency modes involving power synchronization, which increases the interactions between grid-connected converters and DC-link voltages and TAB control.
  

DOCTORAL DEGREE IN ENVIRONMENTAL ENGINEERING

• BULFARO, ISABELLA: Assessment of the Environmental, Economic and Social impacts of Novel Non-Thermal Plasma Methane Cracking process for Hydrogen Production
  
  Author: BULFARO, ISABELLA
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN ENVIRONMENTAL ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 28/10/2026
  Deposit END date: 10/11/2026
  Thesis director: AMANTE GARCIA, BEATRIZ | FERREIRA, VICTOR JOSE
  Thesis abstract: Low-carbon hydrogen is considered a potential option for reducing greenhouse gas (GHG) emissions in hard-to-abate sectors. In this context, methane splitting (MS), also referred to as methane cracking or pyrolysis, has emerged as a potentially lower-carbon hydrogen production route between conventional fossil-based processes and water electrolysis. The first part of this thesis explores the role of MS in Europe in low-carbon hydrogen production; it then examines the technology's sustainability assessment in the scientific literature and the EU regulatory approach to GHG emissions accounting for hydrogen production. The analysis identified key gaps: a lack of sustainability assessments for various MS reactor configurations, inconsistent LCA methodologies, limited co-product end-of-life (EoL) analysis, limited geographical coverage, and no social impact assessment. Given the limitations identified in the review, this research aims to advance knowledge of MS sustainability by analyzing the environmental, economic, and social impacts of the Non-Thermal Methane Splitting (NTP MS) process through a technoeconomic assessment (TEA), a prospective life cycle assessment (pLCA), and a criticality assessment as a socioeconomic indicator.This thesis argues that NTP MS is not an inherently sustainable pathway, recognizing that its sustainability performance depends on technology improvements, co-product management, and regional context. It also proposes new lines of research on emissions associated with carbon use and EoL, aiming to contribute to more robust criteria for the European assessment of MS hydrogen.
  

DOCTORAL DEGREE IN MARINE SCIENCES

• ORIHUELA GARCÍA, Mª ANDREA: Controls on Particulate Organic Carbon Budgets in the North Atlantic Mesopelagic Zone: Process-Based Modelling, Observations, and Model Uncertainty
  
  Author: ORIHUELA GARCÍA, Mª ANDREA
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN MARINE SCIENCES
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 26/10/2026
  Deposit END date: 06/11/2026
  Thesis director: GALÍ TÀPIAS, MARTÍ | RUPRICH ROBERT, YOHAN
  Thesis abstract: This thesis investigates the mechanisms controlling particulate organic carbon (POC) budgets in the ocean, with particular focus on the North Atlantic and the mesopelagic zone. The biological carbon pump transfers organic carbon from the surface ocean to depth, where its residence time is determined by the balance between vertical export, biological transformations and, ultimately, remineralisation and horizontal transport. Despite the critical role of these processes in long-term carbon storage and, hence, in the regulation of global climate, significant uncertainties remain in both observations and their model representation.To address these questions, the thesis combines process-based modelling with a broad set of observational datasets, including satellite products, BGC-Argo floats, gliders, shipboard measurements, and sediment traps. The core modelling framework is the NEMO–PISCES coupled model, employed to diagnose epi- and mesopelagic POC budgets, quantify the relative contribution of different export pathways, and evaluate how detrital size classes, remineralisation rates, and zooplankton-mediated processes shape carbon flux attenuation with depth. Sensitivity simulations explore the influence of key parameters and structural choices, including sinking speed, remineralisation rates, and zooplankton flux feeding. Comparisons with the MEDUSA model and one-dimensional experiments at the Porcupine Abyssal Plain offer a complementary perspective on model uncertainty. In addition, an alternative NEMO–PISCES configuration, forced with phytoplankton-derived surface carbon, is used as a novel approach to assess model uncertainty arising from biases in the representation of upper ocean ecosystem dynamics.The results demonstrate that mesopelagic carbon budgets in PISCES are strongly controlled by the interplay between export composition, particle reactivity, and biological processing within the twilight zone. Small detritus plays a central role in sustaining mesopelagic POC stocks, while zooplankton-mediated transformations and temperature-dependent degradation regulate flux attenuation with depth. The relative importance of each process varies regionally across PISCES: gravitational sinking and microbial remineralisation decline in absolute magnitude yet become progressively more dominant at lower latitudes. The thesis further demonstrates that improvements in surface phytoplankton biomass do not propagate linearly to depth, highlighting the buffering role of the epi- and mesopelagic food webs. Furthermore, case studies reveal that convective and seasonal export events can substantially alter the supply of carbon to depth, and are not consistently captured by standard model parameterisations.Taken together, this thesis provides a mechanistic framework for analysing ocean interior carbon budgets and for identifying structural biases in biogeochemical models. By combining process-based diagnostics with a diverse set of observations, it advances the understanding of carbon export efficiency and remineralisation in the North Atlantic, offering guidance for improving future projections of the ocean carbon cycle and its response to anthropogenic perturbation.
  

DOCTORAL DEGREE IN MECHANICAL, FLUIDS AND AEROSPACE ENGINEERING

• TRAVIESO DISOTUAR, ADRIAN ALBERTO: Enhancing the Surface Integrity and Mechanical Performance of Steels Produced by Laser Powder Bed Fusion through Machining and Vibration-Assisted Ball Burnishing
  
  Author: TRAVIESO DISOTUAR, ADRIAN ALBERTO
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN MECHANICAL, FLUIDS AND AEROSPACE ENGINEERING
  Department: Department of Mechanical Engineering (EM)
  Mode: Normal
  Deposit date: 29/10/2026
  Deposit END date: 11/11/2026
  Thesis director: TRAVIESO RODRIGUEZ, JOSE ANTONIO | JEREZ MESA, RAMON | VILASECA LLOSADA, MONTSERRAT
  Thesis abstract: This dissertation investigates how machining and vibration-assisted ball burnishing (VABB) can improve the surface integrity of metallic components produced by laser powder bed fusion (LPBF). LPBF surfaces commonly exhibit asperities, valleys, attached particles, and other irregularities that may affect functional performance. While machining removes these features through material removal, VABB modifies the surface through localized plastic deformation. This work evaluates both approaches in Maraging 300 steel and 316L stainless steel.The study combines Taguchi L27 optimization with multiscale surface and subsurface characterization using optical profilometry, SEM, Hurst and fractal analysis, EBSD, nanoindentation, X-ray diffraction, and incremental hole drilling. In addition, axial fatigue tests were performed on an independent experimental series of LPBF Maraging 300 steel specimens to evaluate the influence of surface condition on cyclic behavior. All materials were studied without post-build heat treatment.The results show that VABB effectiveness depends on the initial surface condition. As-built surfaces required greater mechanical action, whereas machined surfaces responded effectively to lower forces. Although machining produced the greatest roughness reduction, direct VABB also substantially improved the LPBF topography without completely removing the surface. Reductions in Sa and Sq did not necessarily correspond to greater spatial persistence, hardness, or compressive residual stresses, confirming that surface integrity requires complementary descriptors.In fatigue, machining produced the smoothest surface but did not yield a proportional and uniform increase in fatigue life. Direct VABB showed a tendency toward longer lives at lower stress amplitudes, together with a strong compressive residual stress field, although substantial scatter was also observed in the high cycle region. Overall, LPBF postprocessing should be evaluated through the combined surface and subsurface integrity state rather than roughness alone.
  

DOCTORAL DEGREE IN PHOTONICS

• MEZZADRELLI, ALESSIA: Transparent glass surfaces with functional metallic nanostructures for display applications
  
  Author: MEZZADRELLI, ALESSIA
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN PHOTONICS
  Department: Institute of Photonic Sciences (ICFO)
  Mode: Normal
  Deposit date: 21/10/2026
  Deposit END date: 03/11/2026
  Thesis director: PRUNERI, VALERIO | MAZUMDER, PRANTIK
  Thesis abstract: Display technologies are ubiquitous in everyday life, generally consist of multiple functional components arranged in a stack, where each layer must satisfy specific optical, electrical, mechanical and surface requirements. In this context, nanofabrication offers the possibility to modify material properties and introduce additional functionalities without compromising the primary operation of the device. Glass represents an ideal platform because of its optical transparency, mechanical robustness, chemical stability, and compatibility with thin-film deposition and surface processing. Among the possible material approaches, ultra-thin metal films (UTMFs) are particularly attractive, as their response can be strongly tuned through composition, thickness, morphology, and nanoscale structuring. This thesis investigates the engineering of UTMFs and derived metallic nanostructures on glass for multifunctional display-related applications, with particular emphasis on transparent antimicrobial surfaces and transparent conductors.The first part addresses high-touch surfaces, such as public and shared touchscreens, which can contribute to microbial transmission. Ultra-thin copper films were thermally dewetted to form electrically isolated nanostructures, preserving compatibility with capacitive touch sensing while providing strong antimicrobial activity. Optimization of the thermal treatment enabled control of particle morphology and plasmonic response. Combined with a silica protective layer and an Easy-To-Clean coating, the resulting surfaces achieved bacterial reduction above 99.9%, visible transmission of approximately 70-80%, electrical insulation, and resistance to mechanical abrasion.To further improve the optical performance of antimicrobial surface, Cu-Zn mixed films were investigated as a material-engineering strategy to increase color neutrality while maintaining the antimicrobial functionality of Cu. Co-sputtered Cu-Zn films followed by thermal dewetting produced transparent and electrically insulating nanostructured surfaces with transmission above 80% in the visible range, improved color neutrality, and bacterial reduction exceeding 99.9%. The results demonstrate how compositional engineering can tune the optical response without compromising antimicrobial efficacy.The final part of the thesis explores UTMFs as transparent conductors. Several metals were compared in terms of transmission, sheet resistance, film growth, and percolation behavior, identifying CuZn as a promising material because of its balance between conductivity, transparency, color neutrality, film formation, and time stability. Its optoelectronic properties were further investigated as a function of thickness, thermal annealing, and aging. Compositional analysis indicated a redistribution of Zn toward the surface, associated with the formation of a protective Zn-rich oxidized region and a comparatively Cu-rich metallic layer underneath. This evolution is consistent with the experimentally observed increase in optical transmission and decrease in sheet resistance. Finally, a lithography-free colloidal patterning approach was explored to introduce openings into CuZn UTMF and further tune the trade-off between optical transmission and electrical conductivity.Overall, this work demonstrates how composition, thickness, and nanoscale morphology can be used to engineer ultra-thin metallic systems toward markedly different functionalities. Optical modelling and experimental nanofabrication were combined throughout the work, with particular attention to fabrication routes compatible with scalable industrial processing. The results highlight the potential of UTMFs as a versatile platform for transparent glass surfaces combining optical quality with antimicrobial activity, electrical insulation, or electrical conduction for next-generation display and optoelectronic technologies.
  
• PALARDONIO MENDOZA, SIDNEY: Electrochemical hydrogen storage in graphene oxide and light-induced dehydrogenation of graphane
  
  Author: PALARDONIO MENDOZA, SIDNEY
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN PHOTONICS
  Department: Institute of Photonic Sciences (ICFO)
  Mode: Normal
  Deposit date: 23/10/2026
  Deposit END date: 05/11/2026
  Thesis director: MARTORELL PENA, JORDI | ROS FIGUERAS, CARLES
  Thesis abstract: Renewable electricity is inherently intermittent, and without large-scale storage, it would be difficult to fully displace the use of fossil-based fuels. Hydrogen is a promising energy storage alternative, owing to its high gravimetric energy density. However, its poor volumetric density limits its practical use. Compressed-gas, cryogenic, and metal-hydride approaches address this partially and at a substantial energy penalty. Graphane (C1H1)n, GA), a fully hydrogenated sp3 carbon lattice with a theoretical 7.7 wt% hydrogen capacity in C-H bonds, offers a promising long lasting and safe hydrogen storage. Provided that graphene is chemically inert to hydrogenation, in this thesis we consider a storage strategy based on graphene oxide (GO), whose oxygen functional groups (OFGs) create reactive sites that enable hydrogenation under milder conditions. This thesis examines this GO-to-GA pathway as a single hydrogen storage strategy, constructing an experimental cycle of electrochemical hydrogen incorporation, scaled continuous-flow processing, and thermal or optical hydrogen recovery.The core of the thesis begins in Chapter 2 by examining electrochemical and solar-assisted hydrogenation of GO films. Such electrochemical studies demonstrate that hydrogen incorporation into GO is facilitated by adequately tuning the electrochemical conditions and by the presence of OFGs. We also demonstrate that hydrogenation by periodic cyclic potential modulation outperforms fixed-potential reduction. A performance attributed to continuously exposing fresh OFG sites rather than favouring the competing hydrogen evolution reaction. Coupling a BiVO4/FeNiOOH photoanode with a solar cell shifts the required bias cathodically, indicating hydrogenation may be feasible at substantially reduced voltage. A continuous-flow configuration is explored in Chapter 3 to address upscaling and hydrogenation of GO suspensions. Copper foam outperforms nickel as the cathode material due to its higher overpotential toward H2 evolution, yielding rGO-H at cell potentials as low as 1.5 V with negligible gas evolution.Chapter 4 discusses a photon-based desorption mechanism to probe the C-H desorption dynamics, exploiting selective photon-matched bond destabilization. In this chapter we used GA to better isolate sp3 C-H desorption from O-bound hydrogens present in rGO-H. We demonstrate that 1.55 eV photons (800 nm) dehydrogenate only once supercontinuum generation is reached, causing extensive lattice fragmentation. When using 2.35 eV photons (527 nm), closely matched to the 2.28 eV C-H desorption barrier, we demonstrated dehydrogenation while preserving the carbon framework. This selective removal of the C-H bonds may trigger a self-propagating cascade reaction where the resulting sp2 clusters destabilize and lower the desorption barrier for neighbouring C-H bonds, letting dehydrogenation proceed at a certain point, largely independent of the photon flux.Collectively, these three chapters establish a solar-assisted electrochemical hydrogen loading of GO, while photon-based studies offer mechanistic insight into selective C-H bond breaking and cascade propagation, pointing toward controlled hydrogen release from GA.
  

DOCTORAL DEGREE IN POLYMERS AND BIOPOLYMERS

• AHMED, SAEED: Computational Insights into Conformational Landscapes and Aggregation Propensity of Peptides for Biocompatible Materials
  
  Author: AHMED, SAEED
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN POLYMERS AND BIOPOLYMERS
  Department: Department of Chemical Engineering (EQ)
  Mode: Normal
  Deposit date: 20/10/2026
  Deposit END date: 02/11/2026
  Thesis director: ALEMAN LLANSO, CARLOS ENRIQUE | CONTINI, ALESSANDRO | TORRAS COSTA, JUAN
  Thesis abstract: Protein and peptide aggregation is fundamental to biological self-assembly, supramolecular chemistry, and peptide-based biomaterial design. Non-natural amino acid-containing peptides offer improved structural stability and adaptable conformations, but the molecular factors governing their aggregation remain incompletely understood. This thesis employs a multiscale computational approach combining classical and accelerated molecular dynamics (cMD and aMD), molecular mechanics, and quantum mechanical (QM/DFT) calculations to investigate the conformational landscapes and aggregation behavior of non-natural dipeptides (Cl1 and Cl2) and tetrapeptides (D1Cl1, D1Cl2, D2Cl1, and D2Cl2).Extensive cMD and aMD simulations using the ff19SB force field, complemented by experimental NMR data, were used to characterize peptide folding, stability, flexibility, hydrogen-bonding networks, and dominant conformational states. Structural analyses based on RMSD, radius of gyration, backbone dihedral angles, hydrogen bonds, clustering, and free-energy landscapes revealed prevalent turn motifs and demonstrated that stereochemistry, backbone constraints, and aromatic interactions strongly influence aggregation-prone conformations. Comparison of explicit water and dichloromethane (DCM) showed that solvent environment plays a decisive role in peptide behavior. DCM enhanced intramolecular hydrogen bonding and produced conformational ensembles in closer agreement with experimental NOE distances and J-couplings, whereas water increased solvent competition and conformational dynamics.Intermolecular interactions were further characterized using RMSD, radius of gyration, hydrogen bonds, SASA, π–π stacking, aryl-group geometry, dihedral distances, contact maps, and electrostatic contributions. The results indicate that aggregation is primarily driven by hydrophobic contacts between aryl groups. Enhanced sampling through aMD improved exploration of conformational space and enabled observation of rare aggregation events, metastable intermediates, and alternative assembly pathways that were not accessible through conventional simulations.DFT and SIESTA calculations were used to examine frontier molecular orbitals, HOMO–LUMO gaps, charge distributions, and intermolecular interaction energies. Although electronic-structure calculations provided complementary information on molecular reactivity, stability, and binding, HOMO–LUMO gaps did not statistically distinguish experimentally reported aggregating and non-aggregating peptides and therefore were insufficient as standalone predictors of aggregation. Interaction energies showed relationships with aggregation tendencies observed in molecular dynamics simulations.A computational workflow was also benchmarked using a literature-reported pre-assembled peptide aggregate. Comparison of cMD and aMD simulations demonstrated that enhanced sampling accelerates conformational exploration while preserving the experimentally reported aggregated architecture, supporting the reliability of the computational protocol.Overall, this thesis establishes a systematically validated multiscale framework connecting peptide sequence, stereochemistry, backbone flexibility, solvent environment, hydrogen bonding, aromatic interactions, and electronic structure with peptide folding, stability, self-assembly, and aggregation. The findings demonstrate that minor stereochemical modifications can substantially alter peptide compactness, flexibility, and aggregation propensity, while global electronic descriptors alone cannot capture the complexity of peptide self-assembly. The developed computational framework provides molecular-level insights and supports the rational design of non-natural peptide systems with controlled structural and aggregation properties for peptide-based biomaterials.
  

DOCTORAL DEGREE IN SIGNAL THEORY AND COMMUNICATIONS

• SOLÉ GÓMEZ, JAUME ALEXANDRE: From Structure to Dynamics: Graph Neural Networks for Computational Chemistry
  
  Author: SOLÉ GÓMEZ, JAUME ALEXANDRE
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN SIGNAL THEORY AND COMMUNICATIONS
  Department: Department of Signal Theory and Communications (TSC)
  Mode: Normal
  Deposit date: 27/10/2026
  Deposit END date: 09/11/2026
  Thesis director: RUIZ HIDALGO, JAVIER | RUIZ SABIN, ELISEO
  Thesis abstract: A long-standing aim of chemistry is to connect composition and geometry with measurable properties and with the evolution of systems over time. Experimental measurements provide direct evidence, while computational methods help interpret them and reduce the cost of exploring many systems. Density Functional Theory (DFT) calculations provide widely used reference data, although their computational cost limits large materials searches and long molecular simulations. Machine learning offers a complementary route by learning from examples and evaluating new structures at lower cost. This thesis uses Graph Neural Networks (GNNs), which represent structures as atoms connected by relations and update atomic information through those connections.The first two parts address static-property prediction from periodic crystal structures and relevant conditions without propagating atoms in time. The first develops local and transferable representations and curates an experimental molecular-crystal dataset of anisotropic displacement parameters (ADPs), which are visualised as thermal ellipsoids and describe displacements from average atomic positions, primarily due to thermal motion. CartNet predicts ADPs directly from experimental crystal geometries and measurement temperatures without structural relaxation. This GNN is also evaluated on public scalar crystal-property benchmarks derived from DFT calculations. It also introduces UNATE, which learns atomic-environment representations from unlabelled crystal structures for downstream prediction with limited labelled data.The second part examines whether local crystal graphs are sufficient for static properties that depend on information beyond the local neighbourhood. Some cases require repeated motifs, distant but similar environments, or broader patterns. PRISM addresses this limitation by adding connections to similar atomic environments and coarser views of the periodic crystal. It combines local chemistry with longer-range structural information while preserving periodicity.The final part studies dynamic properties obtained through Molecular Dynamics (MD). Machine Learning Force Fields (MLFFs) supply energies and forces to an MD integrator, which propagates atomic positions and velocities to generate a trajectory. Ensemble averages and time-dependent properties are calculated from this trajectory. This thesis introduces the Multiscale Structural Ensemble (MuSE), a model-agnostic framework that augments a base MLFF with a hierarchy of coarser latent graphs. MuSE constructs this hierarchy through Soft Coarse-Graining Pooling (SCGP), which builds each scale with smooth fractional assignments before message passing gathers information over progressively larger regions. The pooling weights map each scale back to the atoms, and atom-wise gates combine the representations before energy prediction.
  

DOCTORAL DEGREE IN STRUCTURAL ANALYSIS

• DOLZ RIPOLLÉS, MONTSERRAT: NUMERICAL MODELLING AND SIMULATION OF ADDITIVELY MANUFACTURED COMPOSITE MATERIALS. APPLICATION TO KIRIGAMI STRUCTURES
  
  Author: DOLZ RIPOLLÉS, MONTSERRAT
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN STRUCTURAL ANALYSIS
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 30/10/2026
  Deposit END date: 12/11/2026
  Thesis director: MARTÍNEZ GARCÍA, JAVIER | TURON PUJOL, FRANCESC
  Thesis abstract: Additive manufacturing (AM) based on fused-filament processes enables lightweight, customized structural components but exhibits strong anisotropy arising from filament orientation, layer stacking, porosity and interfacial bonding. This thesis develops a predictive and experimentally validated numerical framework to model the stiffness and failure of 3D-printed structures, integrating constitutive modelling, parameter identification, multiscale homogenisation and application-driven analysis. First, a comprehensive review of AM anisotropy, structural applications and kirigami-based deployable systems establishes the motivation, research questions and modelling requirements. The core of the work proposes a finite-element formulation that incorporates imported toolpaths, local material frames, orthotropic constitutive models and elastoplastic formulations, and cohesive modelling of interlayer and interfilament interfaces. A systematic calibration protocol is defined using tensile, shear, bending and cyclic tests, enabling accurate identification of elastic, plastic and fracture parameters. The model is validated on coupons and panels, capturing force-displacement behaviour and stiffness evolution with controlled uncertainty.A multiscale module is introduced to connect micro- and mesostructural descriptors—filament architecture, porosity and interfaces—with effective orthotropic properties through representative volume elements subject to periodic conditions. This enables consistent transfer of material parameters to the macroscale and highlights the role of print orientation on stiffness degradation. Finally, the framework is applied to kirigami-based structures relevant to lightweight deployable systems in naval environments. A dedicated experimental programme (manufacturing, quality control, deployment and compression tests) and its corresponding numerical models allow assessing load capacity, hysteresis, buckling modes and degradation mechanisms. From these results, design rules and charts are derived concerning reinforcement width, print orientation, minimum fold radii and optimal infill patterns. The thesis concludes by quantifying predictive accuracy, outlining methodological and experimental contributions, and identifying limitations and avenues for future research such as high-cycle fatigue, environmental effects and extension to advanced polymer–fibre systems.
  

DOCTORAL DEGREE IN URBANISM

• CATRÓN LAZO, CAROLINA: El azul en el gris: Intervenciones multiescalares en el espacio azul urbano. El caso de las lagunas urbanas de Concepción (Chile).
  
  Author: CATRÓN LAZO, CAROLINA
  Thesis file: (contact the Doctoral School to confirm you have a valid doctoral degree and to get the link to the thesis)
  Programme: DOCTORAL DEGREE IN URBANISM
  Department: Department of Urbanism, Territory and Landscape (DUTP)
  Mode: Change of supervisor
  Deposit date: 23/10/2026
  Deposit END date: 05/11/2026
  Thesis director: GALINDO GONZALEZ, JULIAN | BAERISWYL RADA, SERGIO LEONARDO
  Thesis abstract: This thesis analyses the interventions that influence the relationship between people and urban blue spaces, considering accessibility and visibility as conditions that can expand or restrict opportunities for interaction. The research is conducted through a comparative study of five urban lagoons in Concepción, Chile: Lo Custodio, Lo Galindo, Lo Méndez, Redonda, and Tres Pascualas.The qualitative methodology adopts a multiscalar perspective encompassing the urban, neighbourhood, and specific-site scales. At the urban scale, the relationship between the lagoons and the street network is examined through Space Syntax analysis, complemented by a historical reconstruction. At the neighbourhood scale, potential centralities are identified, together with the barriers and connections that condition the continuity of routes between these centralities and the lagoon access points. At the specific-site scale, the architectural and landscape interventions within the parks are analysed according to their configuration as points, lines, planes, and volumes, as well as their proximity to the water. Visibility is examined through georeferenced photographic records from Instagram.The results show that accessibility is configured through a multiscalar sequence: mobility axes and transfer points at the urban scale; neighbourhood mobility axes and articulation points at the neighbourhood scale; and point, line, plane, and volume interventions at the specific-site scale. Urban interventions connect, neighbourhood interventions articulate, and specific-site interventions distribute. Continuity between these interventions favours progressive movement towards the water, whereas discontinuities restrict it. Visibility has a more localised expression and is primarily associated with interventions that enable positions close to and oriented towards the body of water. Overall, the thesis proposes understanding interventions as a system of relationships across scales, whose articulation conditions opportunities for accessibility and visibility in urban blue spaces.