DeustoTeka
DeustoTeka recoge la producción científica del personal docente e investigador de la Universidad de Deusto. Su propósito es reunir, archivar, preservar y aumentar la visibilidad en acceso abierto de los resultados de investigación.
Comunidades en DSpace
Selecciona una comunidad para explorar sus colecciones.
Envíos recientes
Closed-form resistance and self-inductance of transformer-induced eddy currents in long plates
(Springer Science and Business Media Deutschland GmbH, 2026-08-01) Mauraza Ingelmo, María; Elejabarrieta Olabarri, María Jesus; Cortés Martínez, Fernando
Eddy currents play a key role in a wide range of applications, including induction heating, non-destructive testing, braking, and non-contact damping. The determination of their equivalent electrical parameters, particularly resistance and self-inductance, is essential for physical interpretation and for the validation of numerical models, yet explicit closed-form formulations remain scarce. In this work, closed-form expressions are derived for the resistance and self-inductance associated with transformer-induced eddy currents in conducting plates with large aspect ratios. The analytical formulation enables the evaluation of the induced magnetic field, magnetic flux, and Joule losses, leading to compact expressions for the equivalent electrical parameters and a consistent RL circuit representation of the system. The proposed expressions provide an analytical benchmark for the validation and interpretation of numerical models of transformer-induced eddy currents in long conducting plates. The results are compared with existing analytical approximations obtained under more restrictive assumptions, demonstrating consistency while providing a physically consistent analytical description within the assumptions of dominant longitudinal currents and large aspect ratios.
CFD–DEM simulation of powder flow in a continuous coaxial nozzle for laser metal deposition
(Elsevier B.V., 2026-10-01) Pedrolli, Lorenzo; Achiaga, Beatriz; López, Alejandro
This work presents a fully coupled Computational Fluid Dynamics and Discrete Element Method (CFD–DEM) simulation of powder transport in continuous coaxial Laser Metal Deposition (LMD) nozzles, implemented in Simcenter STAR-CCM+. The methodology integrates a Johnson–Kendall–Roberts (JKR) adhesive contact model via custom field functions to represent cohesive particle–wall interactions influencing residence times and temporal flow structure. A refined polyhedral mesh resolves internal channels and external jet regions while maintaining coupling stability, with physics including two-species gas mixing (argon/air), Gidaspow drag, Sommerfeld shear lift, and gravity. Validation at a representative operating point (3 L/min carrier gas, 15 L/min shielding flow, Rosin–Rammler size distribution, 4.5 g/min feed) yields spatial footprints, centerline peak acceleration of ∼50 m/s2, and a 19% exit mass-flow irregularity (2 ms moving average). The predicted mass-flow irregularity (19% vs. 16% experimental) and accelerations agree closely with measurements, confirming rebound-driven intermittency and azimuthal redistribution at distributor fins as the dominant unsteadiness sources. Conversely, the standoff underprediction (10.2 mm vs. 16 mm) is traced to unresolved exit gas expansion under the incompressible assumption and drag calibration for the particle-laden jet, defining concrete refinement priorities for predictive simulation.
Beyond transparency: governance tensions and a typology of accountability gaps in artificial intelligence for high-stakes service sectors
(Virtus Interpress, 2026) Urrutia Oñate, Dorleta; Bollain Urbieta, Julen; Onieva Caracuel, Enrique; Perallos Ruiz, Asier
Transparency has become the shared reference point for governing artificial intelligence (AI) across regulation, corporate social responsibility (CSR) and explainability in high-stakes services, yet disclosure alone does not produce accountability (Alsulami, 2025; Al Sharif, 2026). This conceptual paper asks where, at which governance level and under which conditions, transparency fails to become accountability, formalising a typology of accountability gaps. The study is a problem-driven synthesis of a bounded corpus of 64 regulatory, organisational and technical sources (2018–2026), coded along two dimensions: governance level (regulatory, organisational, technical or system-level) and accountability condition (formal, distributed or contestable). It identifies four distinct tensions: regulation strengthens formal accountability but underspecifies responsibility and contestation; CSR and ethics frameworks reach distributed accountability when translated into roles and review; Explainable AI (XAI) affords contestability through intelligibility but not recourse; and integrating these levels—where an AI output becomes a reason for action—is the central unresolved gap. Mechanisms at one level cannot substitute for those missing at another: transparency remains necessary but becomes accountability only when connected to responsibility, evidence and contestation. These tensions are amplified in high-stakes service contexts, and the typology offers policymakers, organisations and researchers a diagnostic framework for accountable AI governance.
Enhancing the detection of LTP through lyophilized protein samples and NIR spectroscopy with explainable deep learning
(Nature Research, 2026-12-01) Osa Sánchez, Ainhoa; Del Barrio, Itxasne; Bernardo Seisdedos, Ganeko; Pozo, Sara; García-Zapirain, Begoña
Lipid transfer proteins (LTPs) are clinically relevant allergens widely present in plant-based foods, and their reliable detection in complex food matrices remains a major challenge. In this study, we developed an integrated framework combining near-infrared spectroscopy (NIRS), deep learning, and explainability methods to enable accurate and interpretable identification of LTPs. A total of 11,688 spectral measurements were collect ed using a FLAME-NIR spectrometer (940–1700 nm) from homogenized food samples, purified Pru p 3 and Ara h 9 proteins, and their mixtures with LTP-free matrices such as yogurt and powdered milk. Spectral preprocessing involved first derivative transformation, Standard Normal Variate correction, and feature scaling, followed by dimensionality reduction through a 1D convolutional autoencoder, which generated 64-dimensional latent embeddings. These representations were used to train two deep learning classifiers Convolutional Neural Networks (CNNs) and TabTransformer optimized via Bayesian optimization. The inclusion of purified protein embeddings substantially improved classification performance. The CNN model achieved the highest performance with 95.8% accuracy, 97.3% precision, 96.9% F1-score, and an AUC-ROC of 0.954, outperforming the TabTransformer, which nonetheless reached 95.19% accuracy and 96.4% F1-score. Model explainability was addressed using SHAP and LIME, which identified key latent features corresponding to specific spectral regions (940–1700 nm) associated with allergenic signatures. Compared to baseline models, the protein-enhanced framework demonstrated marked improvements in specificity and overall robustness. These results highlight the value of incorporating purified protein information into AI-based spectral analysis, offering a portable, non-destructive, and interpretable strategy for allergen detection in food safety applications.
A standardised framework for food loss and waste prevention: integrating definitions, causes, destinations, and a circularity index
(Elsevier B.V., 2026-12-01) Amador-Cervera, Manuel; Legarda Macon, Jon; Alonso Vicario, Ainhoa
This paper presents a standardised framework for food loss and waste prevention that harmonises operational definitions, a cause taxonomy, and a hierarchy of destinations aligned with EU regulatory guidance, and demonstrates its application through the full food supply chain case study of a prepared salad in Spain. The framework combines a selection of definitions, methodology to identify points of food loss and waste (FLW) generation using NACE codes, a decision-tree that identifies the feasible optimal destination for each FLW flow, and a FLW circularity index to measure FLW prevention/management performance. Such an index works by weighting destinations (prevention = 1; recycling ≈ 0.80; recovery ≈ 0.55; disposal = 0), with weights chosen via sensitivity analysis of multi-country data and LCA-based evidence. The framework was validated with data covering Primary production, Processing and manufacturing, and Retail and other distribution of food (1 year of data), as well as Households (4 months of data for 10 households), processed with two digital monitoring tools. Results show Processing and manufacturing contributes most to FLW avoided (≈ 60.5%), while Households generate the largest share of FLW (≈ 61.9% of generated flows). The FLW circularity score indicates that the producer achieved the highest efficiency, while households performed the poorest. The results confirm that the framework can be operationalised with existing data sources and digital tools, and is adaptable to diverse contexts and products. The study concludes that this approach provides a robust basis for characterising FLW prevention actions, measuring FLW related metrics, and allowing cross-case comparability.