arXiv subjects

Sergio Hoyas

Publications and source records attributed to Sergio Hoyas.

At least 19 recordsLinked to original sources

arXiv2026

Unveiling wing turbulence dynamics through explainable deep learning

The three-dimensional organization of coherent structures in the turbulent flow over aircraft wings remains elusive, limiting our ability to reduce fuel consumption. Here, we use explainable artificial intelligence to characterize these structures by predicted flow evolution rather than classical predefined kinematic criteria. After training a deep neural network to predict the short-term evolution of the flow, we identify the highest-relevance flow regions using Shapley-value attribution methods. We have found that as the flow decelerates towards the trailing edge, predictive importance shifts from near-wall low-speed structures to three-dimensional pairs of high- and low-speed fluid regions. Matching no single classical structure family, these pairs progressively dominate the dynamically relevant flow. Most stand spanwise side-by-side, enclosing a near-vertical momentum interface with the strongest velocity jump. Their geometry remains invariant in viscous units while their volume expands when approaching separation. Our results reveal a predictive organization that traditional paradigms overlook, opening new ways to control turbulent flows.

physics.flu-dyn↗

arXiv2026

The near-wall cycle for skin-friction generation revealed through explainable deep learning

Skin friction in wall-bounded turbulence is produced by intermittent near-wall motions, yet conventional coherent-structure definitions do not identify which individual events generate wall-shear stress, and thus friction. We train neural networks to predict the future velocity field and wall-shear-stress distribution in turbulent channel flow, and use SHAP-derived importance maps to identify the the input regions most influential for each prediction. The dominant events appear as paired objects: an upstream velocity-relevant region is associated with high-momentum motion toward the wall, while a downstream friction-relevant region marks the enhanced wall-shear-stress left in its wake. Tracking these pairs reveals a recurrent cycle of growth, streamwise elongation, decay, and occasional splitting into new wall-shear-producing events. These results redefine near-wall coherent structures not by what the flow looks like, but by what they do to wall-shear stress.

physics.flu-dyn↗

arXiv2026

AeroJEPA: Learning Semantic Latent Representations for Scalable 3D Aerodynamic Field Modeling

Aerodynamic surrogate models are increasingly used to replace repeated high-fidelity CFD evaluations in many-query design settings, but current approaches still face two important limitations: they often scale poorly to the very large fields arising in realistic 3D aerodynamics, and they rarely produce latent representations that are directly useful for analysis and design. We introduce AeroJEPA, a Joint-Embedding Predictive Architecture for aerodynamic field modeling that addresses both issues. Rather than predicting the full flow field directly from geometry, AeroJEPA predicts a target latent representation of the flow from a context latent representation of the geometry and operating conditions, and optionally reconstructs the field through a continuous implicit decoder. This formulation decouples latent prediction from field resolution while encouraging the latent space to organize semantically. We evaluate AeroJEPA on two complementary datasets: HiLiftAeroML, which stresses the method in a high-fidelity regime with extremely large boundary-layer fields, and SuperWing, which tests large-scale generalization and latent-space optimization over a broad family of transonic wings. Across these benchmarks, AeroJEPA is competitive as a continuous surrogate for aerodynamic fields, scales naturally to high-resolution outputs, and learns context and predicted latents that encode geometry and aerodynamic quantities not used directly as supervision. We further show that the resulting latent space supports controlled interpolation, linear probing, concept-vector arithmetic, and a constrained design latent-optimization experiment. These results suggest that predictive latent learning is a promising direction for scalable and design-meaningful aerodynamic surrogate modeling.

cs.LG↗

arXiv2026

Agentic Exploration of PDE Spaces using Latent Foundation Models for Parameterized Simulations

Flow physics and more broadly physical phenomena governed by partial differential equations (PDEs), are inherently continuous, high-dimensional and often chaotic in nature. Traditionally, researchers have explored these rich spatiotemporal PDE solution spaces using laboratory experiments and/or computationally expensive numerical simulations. This severely limits automated and large-scale exploration, unlike domains such as drug discovery or materials science, where discrete, tokenizable representations naturally interface with large language models. We address this by coupling multi-agent LLMs with latent foundation models (LFMs), a generative model over parametrised simulations, that learns explicit, compact and disentangled latent representations of flow fields, enabling continuous exploration across governing PDE parameters and boundary conditions. The LFM serves as an on-demand surrogate simulator, allowing agents to query arbitrary parameter configurations at negligible cost. A hierarchical agent architecture orchestrates exploration through a closed loop of hypothesis, experimentation, analysis and verification, with a tool-modular interface requiring no user support. Applied to flow past tandem cylinders at Re = 500, the framework autonomously evaluates over 1,600 parameter-location pairs and discovers divergent scaling laws: a regime-dependent two-mode structure for minimum displacement thickness and a robust linear scaling for maximum momentum thickness, with both landscapes exhibiting a dual-extrema structure that emerges at the near-wake to co-shedding regime transition. The coupling of the learned physical representations with agentic reasoning establishes a general paradigm for automated scientific discovery in PDE-governed systems.

cs.AI↗

arXiv2026

Explainable deep learning reveals the physical mechanisms behind the turbulent kinetic energy equation

In this work, we investigate the physical mechanisms governing turbulent kinetic energy transport using explainable deep learning (XDL). An XDL model based on SHapley Additive exPlanations (SHAP) is used to identify and percolate high-importance structures for the evolution of the turbulent kinetic energy budget terms of a turbulent channel flow at a friction Reynolds number of $Re_τ= 125$. The results show that the important structures are predominantly located in the near-wall region and are more frequently associated with sweep-type events. In the viscous layer, the SHAP structures relevant for production and viscous diffusion are almost entirely contained within those relevant for dissipation, revealing a clear hierarchical organization of near-wall turbulence. In the outer layer, this hierarchical organization breaks down and only velocity-pressure-gradient correlation and turbulent transport SHAP structures remain, with a moderate spatial coincidence of approximately $60\%$. Finally, we show that none of the coherent structures classically studied in turbulence are capable of representing the mechanisms behind the various terms of the turbulent kinetic energy budget throughout the channel. These results reveal dissipation as the dominant organizing mechanism of near-wall turbulence, constraining production and viscous diffusion within a single structural hierarchy that breaks down in the outer layer.

physics.flu-dyn↗

arXiv2025

Classically studied coherent structures only paint a partial picture of wall-bounded turbulence

For the last 140 years, the mechanisms of transport and dissipation of energy in a turbulent flow have not been completely understood. Previous research has focused on analyzing the so-called coherent structures, organized flow patterns characterized by their spatial coherence, lifespan and significant contribution to momentum and energy transfer. However, the connection between these structures and the flow development is still uncertain. Here, we show a data-driven methodology for objectively identifying high-importance regions. A deep-learning model is trained to predict a future state of the flow and the gradient-SHAP explainability algorithm is used to calculate the importance of each grid point. Finally, high-importance regions are computed using the SHAP data and are compared to the other coherent structures. The SHAP analysis provides an objective way to identify the regions of higher importance, which exhibit different levels of agreement with the classical structures without being completely related to any particular one.

physics.flu-dyn↗

arXiv2025

Exploring the interplay between Planetary Boundaries and Sustainable Development Goals using Large Language Models

By analyzing 40,037 climate articles using Large Language Models (LLMs), we identified interactions between Planetary Boundaries (PBs) and Sustainable Development Goals (SDGs). An automated reasoner distinguished true trade-offs (SDG progress harming PBs) and synergies (mutual reinforcement) from double positives and negatives (shared drivers). Results show 21.1% true trade-offs, 28.3% synergies, and 19.5% neutral interactions, with the remainder being double positive or negative. Key findings include conflicts between land-use goals (SDG2/SDG6) and land system boundaries (PB6), together with the underrepresentation of social SDGs in the climate literature. Our study highlights the need for integrated policies that align development goals with planetary limits to reduce systemic conflicts. We propose three steps: (1) integrated socio-ecological metrics, (2) governance ensuring that SDG progress respects Earth system limits, and (3) equity measures protecting marginalized groups from boundary compliance costs.

cs.CY↗

arXiv2025

Evaluating Visual Mathematics in Multimodal LLMs: A Multilingual Benchmark Based on the Kangaroo Tests

Multimodal Large Language Models (MLLMs) promise advanced vision language capabilities, yet their effectiveness in visually presented mathematics remains underexplored. This paper analyzes the development and evaluation of MLLMs for mathematical problem solving, focusing on diagrams, multilingual text, and symbolic notation. We then assess several models, including GPT 4o, Pixtral, Qwen VL, Llama 3.2 Vision variants, and Gemini 2.0 Flash in a multilingual Kangaroo style benchmark spanning English, French, Spanish, and Catalan. Our experiments reveal four key findings. First, overall precision remains moderate across geometry, visual algebra, logic, patterns, and combinatorics: no single model excels in every topic. Second, while most models see improved accuracy with questions that do not have images, the gain is often limited; performance for some remains nearly unchanged without visual input, indicating underutilization of diagrammatic information. Third, substantial variation exists across languages and difficulty levels: models frequently handle easier items but struggle with advanced geometry and combinatorial reasoning. Notably, Gemini 2.0 Flash achieves the highest precision on image based tasks, followed by Qwen VL 2.5 72B and GPT 4o, though none approach human level performance. Fourth, a complementary analysis aimed at distinguishing whether models reason or simply recite reveals that Gemini and GPT 4o stand out for their structured reasoning and consistent accuracy. In contrast, Pixtral and Llama exhibit less consistent reasoning, often defaulting to heuristics or randomness when unable to align their outputs with the given answer options.

cs.AI↗

arXiv2025

Large language models in climate and sustainability policy: limits and opportunities

As multiple crises threaten the sustainability of our societies and pose at risk the planetary boundaries, complex challenges require timely, updated, and usable information. Natural-language processing (NLP) tools enhance and expand data collection and processing and knowledge utilization capabilities to support the definition of an inclusive, sustainable future. In this work, we apply different NLP techniques, tools and approaches to climate and sustainability documents to derive policy-relevant and actionable measures. We focus on general and domain-specific large language models (LLMs) using a combination of static and prompt-based methods. We find that the use of LLMs is successful at processing, classifying and summarizing heterogeneous text-based data. However, we also encounter challenges related to human intervention across different workflow stages and knowledge utilization for policy processes. Our work presents a critical but empirically grounded application of LLMs to complex policy problems and suggests avenues to further expand Artificial Intelligence-powered computational social sciences.

cs.CY↗

arXiv2024

Additive-feature-attribution methods: a review on explainable artificial intelligence for fluid dynamics and heat transfer

The use of data-driven methods in fluid mechanics has surged dramatically in recent years due to their capacity to adapt to the complex and multi-scale nature of turbulent flows, as well as to detect patterns in large-scale simulations or experimental tests. In order to interpret the relationships generated in the models during the training process, numerical attributions need to be assigned to the input features. One important example are the additive-feature-attribution methods. These explainability methods link the input features with the model prediction, providing an interpretation based on a linear formulation of the models. The SHapley Additive exPlanations (SHAP values) are formulated as the only possible interpretation that offers a unique solution for understanding the model. In this manuscript, the additive-feature-attribution methods are presented, showing four common implementations in the literature: kernel SHAP, tree SHAP, gradient SHAP, and deep SHAP. Then, the main applications of the additive-feature-attribution methods are introduced, dividing them into three main groups: turbulence modeling, fluid-mechanics fundamentals, and applied problems in fluid dynamics and heat transfer. This review shows thatexplainability techniques, and in particular additive-feature-attribution methods, are crucial for implementing interpretable and physics-compliant deep-learning models in the fluid-mechanics field.

physics.flu-dyn↗

arXiv2024

Sensitivity study of resolution and convergence requirements for extended overlap region in wall-bounded turbulence

Direct Numerical Simulations (DNSs) are one of the most powerful tools for studying turbulent flows. Even if achievable Reynolds numbers are lower than those obtained with experimental means, there is a clear advantage since the entire velocity field is known, and any desired quantity can be evaluated. This also includes the computation of derivatives of all relevant terms. One such derivative provides the indicator function, which is the product of wall distance by wall-normal derivative of the mean streamwise velocity. This derivative may depend on mesh spacing and distribution. However, it is extremely affected by the convergence of the simulation. The indicator function is a cornerstone to understanding inner and outer interactions in wall-bounded flows and describing the overlap region between them. We find a clear dependence of this indicator function on mesh distributions we examined, raising questions about classical mesh and convergence requirements for DNS and achievable accuracy. Within the framework of the logarithmic plus linear overlap region, coupled with a parametric study of channel flows and some pipe flows, sensitivities of extracted overlap parameters are examined, and a path is revealed to establishing their high-$Re_τ$ or near-asymptotic values at modest Reynolds numbers accessible by high-quality DNS with reasonable ``cost''.

physics.flu-dyn↗

arXiv2024

Higher DNS-resolution requirements for expanded overlap region and confirmation of a convergence criterion

Direct numerical simulations (DNS) stand out as formidable tools in studying turbulent flows. Despite the fact that the achievable Reynolds number remains lower than those available through experimental methods, DNS offers a distinct advantage: the complete knowledge of the velocity field, facilitating the evaluation of any desired quantity. This capability should extend to compute derivatives. Among the classic functions requiring derivatives is the indicator function, $Ξ(y^+) = y^+\frac{{\rm d}\overline{U}_x^+}{{\rm d}y^+}$. This function encapsulates the wall-normal derivative of the streamwise velocity, with its value possibly influenced by mesh size and its spatial distribution. The indicator function serves as a fundamental element in unraveling the interplay between inner and outer layers in wall-bounded flows, including the overlap region, and forms a critical underpinning in turbulence modeling. Our investigation reveals a sensitivity of the indicator function on the utilized mesh distributions, prompting inquiries into the conventional mesh-sizing paradigms for DNS applications.

physics.flu-dyn↗

arXiv2024

Utilizing indicator functions with computational data to confirm nature of overlap in normal turbulent stresses: logarithmic or quarter-power

Indicator functions of the streamwise normal-stress profiles (NSP), based on careful differentiation of some of the best direct numerical simulations (DNS) data from channel and pipe flows, over the range $550<Re_τ<16,000$, are examined to establish the existence and range in wall distances of either a logarithmic-trend segment or a $1/4$-power region. For the nine out of fifteen cases of DNS data we examined where $Re_τ<2,000$, the NSP did not contain either of the proposed trends. As $Re_τ$ exceeds around $2,000$ a $1/4$-power, reflecting the ``bounded-dissipation'' predictions of Chen \& Sreenivasan and data analysis of Monkewitz , develops near $y^+=1,000$ and expands with Reynolds numbers extending to $1,000<y^+<10,000$ for $Re_τ$ around $15,000$. This range of $1/4$-power NSP corresponds to a range of outer-scaled $Y$ between around $0.3$ and $0.7$. The computational database examined did not include the zero-pressure-gradient boundary layer experiments at higher Reynolds numbers where the logarithmic trend in the NSP has been previously reported around $y^+$ of $1,000$ by Marusic et al. according to a ``wall-scaled eddy model''.

physics.flu-dyn↗

arXiv2024

Identifying regions of importance in wall-bounded turbulence through explainable deep learning

Despite its great scientific and technological importance, wall-bounded turbulence is an unresolved problem in classical physics that requires new perspectives to be tackled. One of the key strategies has been to study interactions among the energy-containing coherent structures in the flow. Such interactions are explored in this study for the first time using an explainable deep-learning method. The instantaneous velocity field obtained from a turbulent channel flow simulation is used to predict the velocity field in time through a U-net architecture. Based on the predicted flow, we assess the importance of each structure for this prediction using the game-theoretic algorithm of SHapley Additive exPlanations (SHAP). This work provides results in agreement with previous observations in the literature and extends them by revealing that the most important structures in the flow are not necessarily the ones with the highest contribution to the Reynolds shear stress. We also apply the method to an experimental database, where we can identify completely new structures based on their importance score. This framework has the potential to shed light on numerous fundamental phenomena of wall-bounded turbulence, including novel strategies for flow control.

physics.flu-dyn↗

arXiv2024

Validation of symmetry-induced high moment velocity and temperature scaling laws in a turbulent channel flow

The symmetry-based turbulence theory has been used to derive new scaling laws for the streamwise velocity and temperature moments of arbitrary order. For this, it has been applied to an incompressible turbulent channel flow driven by a pressure gradient with a passive scalar equation coupled in. To derive the scaling laws, symmetries of the classical Navier-Stokes and the thermal energy equations have been used together with statistical symmetries, i.e. the statistical scaling and translation symmetries of the multi-point moment equations. Specifically, the multi-point moments are built on the instantaneous velocity and temperature fields other than in the classical approach, where moments are based on the fluctuations of these fields. With this instantaneous approach, a linear system of multi-point correlation equations has been obtained, which greatly simplifies the symmetry analysis. The scaling laws have been derived in the limit of zero viscosity and heat conduction, i.e. $Re_τ\rightarrow \infty$ and $Pr > 1$, and apply in the centre of the channel, i.e. they represent a generalization of the deficit law so herewith extending the work of done on the velocity field. The scaling laws are all power laws, with the exponent of the high moments all depending exclusively on those of the first and second moments. To validate the new scaling laws, the data from a large number of DNS for different Reynolds and Prandtl numbers have been used. The results show a very high accuracy of the scaling laws to represent the DNS data. The statistical scaling symmetry of the multi-point moment equations, which characterizes intermittency, has been the key to the new results since it generates a constant in the exponent of the final scaling law. Most important, since this constant is independent of the order of the moments, it clearly indicates anomalous scaling.

math-ph↗

arXiv2023

Data-driven assessment of arch vortices in simplified urban flows

Understanding flow structures in urban areas is widely recognized as a challenging concern due to its effect on urban development, air quality, and pollutant dispersion. In this study, state-of-the-art data-driven methods for modal analysis of simplified urban flows are used to study the dominant flow processes in these environments. Higher order dynamic mode decomposition (HODMD), a highly-efficient method to analyze turbulent flows, is used together with traditional techniques such as proper-orthogonal decomposition (POD) to analyze high-fidelity simulation data of a simplified urban environment. Furthermore, the spatio-temporal Koopman decomposition (STKD) will be applied to the temporal modes obtained with HODMD to perform spatial analysis. The flow interaction within the canopy influences the flow structures, particularly the arch vortex. The latter is a vortical structure generally found downstream of wall-mounted obstacles, which is generated as a consequence of flow separation. Therefore, the main objective of the present study is to characterize the mechanisms that promote these phenomena in urban areas with different geometries. Remarkably, among all the vortical structures identified by the HODMD algorithm, low- and high-frequency modes are classified according to their relation with the arch vortex. They are referred to as vortex-generator and vortex-breaker modes, respectively. This classification implies that one of the processes driving the formation and destruction of major vortical structures in between the buildings is the interaction between low- and high-frequency structures. The high energy revealed by the POD for the vortex-breaker modes points to this destruction process as the mechanism driving the flow dynamics.

physics.flu-dyn↗

arXiv2022

The Sustainable Development Goals and Aerospace Engineering: A critical note through Artificial Intelligence

The 2030 Agenda of the United Nations (UN) revolves around the Sustainable Development Goals (SDGs). A critical step towards that objective is identifying whether scientific production aligns with the SDGs' achievement. To assess this, funders and research managers need to manually estimate the impact of their funding agenda on the SDGs, focusing on accuracy, scalability, and objectiveness. With this objective in mind, in this work, we develop ASDG, an easy-to-use artificial-intelligence (AI)-based model for automatically identifying the potential impact of scientific papers on the UN SDGs. As a demonstrator of ASDG, we analyze the alignment of recent aerospace publications with the SDGs. The Aerospace data set analyzed in this paper consists of approximately 820,000 papers published in English from 2011 to 2020 and indexed in the Scopus database. The most-contributed SDGs are 7 (on clean energy), 9 (on industry), 11 (on sustainable cities) and 13 (on climate action). The establishment of the SDGs by the UN in the middle of the 2010 decade did not significantly affect the data. However, we find clear discrepancies among countries, likely indicative of different priorities. Also, different trends can be seen in the most and least cited papers, with clear differences in some SDGs. Finally, the number of abstracts the code cannot identify is decreasing with time, possibly showing the scientific community's awareness of SDG.

cs.DL↗

arXiv2022

A deep-learning approach for reconstructing 3D turbulent flows from 2D observation data

Turbulence is a complex phenomenon that has a chaotic nature with multiple spatio-temporal scales, making predictions of turbulent flows a challenging topic. Nowadays, an abundance of high-fidelity databases can be generated by experimental measurements and numerical simulations, but obtaining such accurate data in full-scale applications is currently not possible. This motivates utilising deep learning on subsets of the available data to reduce the required cost of reconstructing the full flow in such full-scale applications. Here, we develop a generative-adversarial-network (GAN)-based model to reconstruct the three-dimensional velocity fields from flow data represented by a cross-plane of unpaired two-dimensional velocity observations. The model could successfully reconstruct the flow fields with accurate flow structures, statistics and spectra. The results indicate that our model can be successfully utilised for reconstructing three-dimensional flows from two-dimensional experimental measurements. Consequently, a remarkable reduction of the experimental setup cost can be achieved.

physics.flu-dyn↗