Pasar al contenido principal

ANALISIS IMPULSADO POR INTELIGENCIA ARTIFICIAL DE ESTRUCTURAS TURBULENTAS COHERENTES: UN CAMINO HACIA LA REDUCCION DE LA RESISTENCIA EN PERFILES ALARES

PID2024-162480OB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para la Investigación y el Desarrollo Experimental
Subprograma Subprograma Estatal de Generación de Conocimiento Científico-Técnico y Desarrollo Experimental
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2024
Plan Estatal 2024-2027 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSITAT POLITÈCNICA DE VALÈNCIA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 5
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Spatial and Energetic Organization of Coherent Structures in Couette-Poiseuille Turbulent Channels

  • Gandía-Barberá, Sergio
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] Coherent structures play a pivotal role in wall-bounded turbulence, serving as primary carriers of momentum, energy, and scalar quantities across the flow. This study examines coherent structures, specifically streamwise streaks and intense Reynolds stress regions (Q structures), within a novel DNS dataset capturing a stepped transition from pure Poiseuille flow to pure Couette flow at Re tau approximate to 250, based on the stationary wall. Structures are identified using a percolation algorithm to ensure well-defined boundaries, followed by three-dimensional clustering in Cartesian coordinates. They are further classified as wall-attached or wall-detached based on their proximity to the domain walls. Intense Reynolds stress structures are categorized into quadrants according to the signs of their averaged velocity components. The statistical properties of these structures-encompassing geometric characteristics, energy content, and spatial distribution-are thoroughly analyzed. Particular emphasis is placed on how these properties evolve across the transition from Poiseuille to Couette flow. The results reveal that increasing mean shear in Couette-like cases significantly influences the energy content and spatial distribution of the structures while their geometric characteristics remain relatively consistent across the dataset. This spatial distribution is closely linked to the large-scale structures of the streamwise velocity component in Couette flow, confirming that these structures are genuine physical features rather than artificial artifacts of the flow., The authors acknowledge support by CIN/AEI/10.13039/501100011033 and by ERDF, "A way of making Europe", under project PID2024-162480OB-I00.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

From Black-Box Optimization to Importance-Guided Control: A Perspective on Explainable Deep Reinforcement Learning for Drag Reduction

  • Reverte-Badillo, Belén
  • Vinuesa, Ricardo
  • Clara Trillo-Yagüe
  • Cremades-Botella, Andrés|||0000-0002-7052-4913
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] Fluid-dynamic drag accounts for a substantial fraction of energy consumption across air, ground, and maritime transport systems, making its reduction a critical lever for decarbonizing mobility. While active flow control (AFC) strategies have demonstrated significant drag reduction potential, their design remains constrained by heuristic physical assumptions about dominant flow structures. Recent developments in deep reinforcement learning (DRL) have emerged as a transformative paradigm, capable of autonomously discovering control strategies in high-dimensional turbulent environments. This perspective traces the evolution of drag reduction approaches from classical passive and active control approaches toward data-driven methods based on DRL. A particularly promising direction is the integration of explainable artificial intelligence (XAI) with DRL, which provides physically interpretable information about flow regions associated with drag generation and guides the learning process toward physically meaningful actuation schemes. As a result, XAI-guided DRL controllers have been shown in canonical configurations to achieve comparable or improved drag reduction with substantially lower actuation power than controllers trained directly for drag minimization. This transition from opaque optimization toward flow control informed by dynamical causal relationships represents a key step for the development of energy-efficient and sustainable flow-control solutions for transport systems., SH and BR acknowledge support by CIN/AEI/10.13039/501100011033 and by ERDF, A way of making Europe , under project PID2024-162480OB-I00 and CIPROM/2024/35 by GVA.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Generalized Chi distribution for non-integer degrees of freedom in modelling biological collectivities

  • Castro-Palacio, Juan Carlos|||0000-0002-0132-9989
  • Fernández de Córdoba, Pedro|||0000-0002-0347-7280
  • González-Cortés, Álvaro|||0009-0001-2328-2173
  • Isidro, J.M.|||0000-0002-0720-9945
  • Noverques-Medina, Aina|||0000-0002-7810-0381
  • Sánchez Pérez, Enrique Alfonso|||0000-0001-8854-3154
  • Orellana-Panchame, Marcos Fabricio
  • Sahu, Saira
[EN] This article aims to expand the practical potential of the Chi and Chi-square distributions by extending their applicability to non-integer degrees of freedom, k. The study has two main objectives: first, to explore and leverage the relationship between the Chi, Chi-square, and Gamma distributions to interpret non-integer degrees of freedom; and second, to apply this framework to real survival-time data in the context of biological and health sciences. Three specific cases are analyzed: the germination times of Pinus tropicalis Morelet seeds cultivated in western Cuba; the interval between infection by virulent tuberculous bacilli and death from tuberculosis in guinea pigs; and the period from the onset of symptoms to death due to COVID-19 in patients in Mexico during 2020. A parameter estimation was performed by maximum likelihood (MLE), and the optimal value of k was found to be non-integer in all cases., We would like to acknowledge funding from the Generalitat Valenciana (Spain) through the PROMETEO CIPROM/2023/32 grant, and from the Agencia Estatal de Investigacion (Spain) through ¿the PID2022-142407NB-I00 and PID2024-162480OB-I00 grants. The work of S.S. is partially supported by DGAPA-UNAM (Mexico) Project No. IN103522. The author A.G.C. was supported ¿ by a contract of the Programa de Ayudas de Investigacion y Desarrollo (PAID-01-24), Universitat ¿Politecnica de Val ` encia
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

A Comparison of Supervised Machine Learning Algorithms for Defective Pixel Detection

  • López-Portilla, Bárbaro M.
  • Balzer, Kristian
  • Carballo, Lorena
  • Figueroa, Miguel
  • Peset Mancebo, María Fernanda|||0000-0003-3706-6532
  • Iglesias-Martínez, Miguel Enrique|||0000-0003-1405-5099
  • Fernández de Córdoba, Pedro|||0000-0002-0347-7280
[EN] Detecting defective pixels in CMOS Image Sensors is a critical task for ensuring high-quality image acquisition, as even minor defects can significantly degrade performance in vision systems and downstream processing. This study evaluates the performance of several supervised machine learning algorithms for defective pixel detection using 300 grayscale images of 512 × 512 pixels from the publicly available TAMPERE17 dataset, in which dead and hot pixels were randomly introduced. The evaluated algorithms include k-Nearest Neighbors (kNN), Naïve Bayes, Decision Tree, Random Forest, Support Vector Machine (SVM), and Multilayer Perceptron. Performance was assessed using specificity, recall, precision, and the phi coefficient. A sensitivity analysis was additionally performed using neighborhood windows of 3 × 3, 5 × 5, and 7 × 7 pixels during feature extraction. The results showed that the neighborhood size has a noticeable impact on classification performance, with the best overall results obtained using an SVM classifier with an RBF kernel and a 3 × 3 feature extraction window. This configuration achieved a phi coefficient of 98.08%, together with a specificity of 99.00%, a recall of 99.00%, and a precision of 99.02%. Statistical analysis further indicated that SVM (RBF) and Multilayer Perceptron exhibited statistically comparable performance under the evaluated experimental conditions. The analysis further revealed that defective pixels are primarily characterized by highly local intensity variations, while larger neighborhoods do not necessarily improve classification performance. Compared with previously reported defective pixel detection methods, the proposed approach achieved highly competitive results and the highest phi coefficient among the evaluated methods. Additionally, the impact of defective pixel detection on image quality was assessed using a simple median filter correction stage and evaluated through Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). The best-performing configuration achieved a PSNR of 40.96 dB and an SSIM of 98.68% after median filter correction, demonstrating substantial improvement over median filtering alone, although the resulting image quality remains below that of the best-performing state-of-the-art correction methods. The proposed framework provides reproducible reference results for evaluating supervised machine learning approaches to defective pixel detection while analyzing the influence of feature extraction window size on classification performance and reconstructed image quality, The authors would like to thank the Departamento de Procesos Industriales, Facultad de Ingeniería, Universidad Católica de Temuco, for providing the infrastructure and institutional support that made this work possible. Miguel Figueroa acknowledges the support of the Chilean National Agency for Research and Development (ANID) through Centers for Applied Research CIA250006 and Specific Thematic Research Team Projects ATE250003. Pedro Fernández de Córdoba gratefully acknowledges the support of the Generalitat Valenciana (Spain) through the PROMETEO grant CIPROM/2023/32 and the Agencia Estatal de Investigación (Spain) through projects PID2022-142407NB-I00 and PID2024-162480OB-I00. Miguel E. Iglesias Martínez s research was supported by the postdoctoral research grant Subvenciones para la contratación de personal investigador en fase postdoctoral APOSTD/2025 (MODALIDAD A), funded by the Conselleria de Educación, Cultura, Universidades y Empleo of the Generalitat Valenciana (Reference CIAPOS/2024/238) and co-funded by the European Social Fund Plus (FSE+) through the 2021 2027 Operational Program of the Comunitat Valenciana.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Comparison between wall and dissipative scalings for fourth-order structure functions in a fully developed turbulent channel flow

  • Liu, D.
  • Antonia, R. A.
  • Tang, S. L.
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] The classical scaling of wall turbulence is based on wall parameters (friction velocity, friction temperature, and kinematic viscosity). Recently, a novel scaling based on dissipative scales-the mean turbulent energy dissipation rate, the mean scalar dissipation rate, and kinematic viscosity-was put forward by Tang and Antonia ["Scaling of small-scale wall turbulence," J. Fluid Mech. 948, A25 (2022) and "Similarity for dissipation-scaled wall turbulence," J. Fluid Mech. 960, A18 (2023)]. This new scaling is supported by extensive wall turbulence data, related mainly to the second- and third-order statistics. In this paper, we systematically compare the classical wall scaling and the dissipative scaling in the context of fourth-order statistics, specifically, the fourth-order velocity and scalar structure functions in a fully developed turbulent channel flow at a molecular Prandtl number Pr=0.71. It is found that both the wall parameter- and dissipative scale-normalized fourth-order velocity and scalar structure functions exhibit similar degrees of collapse at moderate distances from the wall. However, in the very-near-wall region, dissipative scale-normalized structure functions collapse approximately at all scales, while wall parameter-normalized distributions vary systematically with the Reynolds number. Far from the wall, dissipative scale-normalized structure functions collapse approximately at small scales, while wall parameter-normalized distributions continue to evolve with the Reynolds number at all scales. Overall, the scaling based on dissipative scales is superior to that based on wall parameters over a significant portion of the channel, except at moderate distances from the wall, where both scalings are valid in the small-scale range., S.L.T. wishes to acknowledge the support given to him from the National Natural Science Foundation of China (Project No. 12572321), Guangdong Basic and Applied Basic Research Foundation (Project No. 2023B1515020069) and Shenzhen Science and Technology Program (Project Nos. RCYX20210706092046085 and GXWD20220817171516009). S.H. was funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe," under project PID2024-162480OB-I00. D.L. wishes to acknowledge Mr. Xinjun Li's help in the calculation of the fourth-order structure functions in channel flow.