Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 17
Encontrada(s) 1 página(s)

e-Archivo. Repositorio Institucional de la Universidad Carlos III de Madrid
Publicación Artículo científico (JournalArticle)

β-Variational autoencoders and transformers for reduced-order modelling of fluid flows

  • Solera Rico, Alberto
  • Sanmiguel Vila, Carlos
  • Gómez-López, Miguel
  • Wang, Yuning
  • Almashjary, Abdulrahman
  • Dawson, Scott T.M.
  • Vinuesa, Ricardo
Variational autoencoder architectures have the potential to develop reduced-order models for chaotic fluid flows. We propose a method for learning compact and near-orthogonal reduced-order models using a combination of a ß-variational autoencoder and a transformer, tested on numerical data from a two-dimensional viscous flow in both periodic and chaotic regimes. The ß-variational autoencoder is trained to learn a compact latent representation of the flow velocity, and the transformer is trained to predict the temporal dynamics in latent-space. Using the ß-variational autoencoder to learn disentangled representations in latent-space, we obtain a more interpretable flow model with features that resemble those observed in the proper orthogonal decomposition, but with a more efficient representation. Using Poincaré maps, the results show that our method can capture the underlying dynamics of the flow outperforming other prediction models. The proposed method has potential applications in other fields such as weather forecasting, structural dynamics or biomedical engineering., RV acknowledges the financial support from ERC grant no. ‘2021-CoG-101043998, DEEPCONTROL’. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
e-Archivo. Repositorio Institucional de la Universidad Carlos III de Madrid
Publicación Artículo científico (JournalArticle)

Towards optimal β-variational autoencoders combined with transformers for reduced-order modelling of turbulent flows

  • Wang, Yuning
  • Solera Rico, Alberto
  • Sanmiguel Vila, Carlos
  • Vinuesa, Ricardo
Variational autoencoders (VAEs) have shown promising potential as artificial neural networks (NN) for developing reduced-order models (ROMs) in the context of turbulent flows. In this study, we propose a method that combines beta-VAEs for modal decomposition and transformer neural networks for temporal-dynamics prediction in the latent space to develop ROMs. We apply our method to an existing database of a turbulent flow around a wall-mounted square cylinder obtained by direct numerical simulation (DNS). A parametric study is performed to investigate the effects of the hyperparameters of the proposed beta-VAEs and determine the optimal values. For the first time, we incorporate the consideration of the complexity of architecture into our studies, providing new insights into hyperparameter selection for beta-VAEs, which remains a challenging problem for optimising model performance. Results regarding the influence of the different hyperparameters and guidelines to design these architectures are reported. Our optimal model achieves a reconstruction accuracy of 97.18% of the entire dataset using only ten modes. Subsequently, we employ the transformer models to identify latent-space temporal dynamics learned by the optimal beta-VAE model and build ROMs to predict instantaneous fields. The resulting model achieves promising accuracy in temporal-dynamics predictions and yields energy reconstruction levels of 96.5% and 83% for a field 25 and 50 steps into the future, respectively, showcasing the potential of the transformer in predicting the temporal dynamics. Overall, the proposed method has potential applications in advanced flow control and fundamental studies of complex turbulent flows., Part of the deep-learning-model training was enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) at Berzelius (NSC), partially funded by the Swedish Research Council through grant agreement no. 2022-06725. RV acknowledges the financial support from ERC grant no. ‘2021-CoG-101043998, DEEPCONTROL’. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

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

  • Hoyas, Sergio|||0000-0002-8458-7288
  • Vinuesa, Ricardo
  • Schmid, Peter
  • Nagib, Hassan
[EN] Direct numerical simulations (DNSs) are among the most powerful tools for studying turbulent flows. Even though the achievable Reynolds numbers are lower than those obtained through experimental means, DNS offers a clear advantage: The entire velocity field is known, allowing for the evaluation of any desired quantity. This capability includes the computation of derivatives of all relevant terms. One such derivative provides the indicator function, which is the product of the wall distance and the wall-normal derivative of the mean streamwise velocity. This derivative may depend on mesh spacing and distribution, but it is extremely affected by the convergence of the simulation. The indicator function is crucial for 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 the mesh distributions we examine, 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. This study reveals a path to establishing their high-Rey y or nearly asymptotic values at modest Reynolds numbers, but larger than the ones used in this work, accessible by high-quality DNS with reasonable cost., Computer time was provided by King Abdullah University of Science and Technology, Project No. K1652. R.V. was supported by ERC Grant No. 2021-CoG-101043998, DEEPCONTROL. The views and opinions expressed are however those of the author (s) only and do not necessarily reflect those of European Union or European Research Council. S.H. was funded by Project No. PID2021-128676OB-I00 from Ministerio de Ciencia, Innovacion y Universidades/FEDER.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

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

  • Cremades-Botella, Andrés|||0000-0002-7052-4913
  • Hoyas, Sergio|||0000-0002-8458-7288
  • Quintero-Igeño, Pedro-Manuel|||0000-0003-4373-2079
  • Deshpande, Rahul
  • Lellep, Martin
  • Lee, Will Junghoon
  • Monty, Jason P.
  • Hutchins, N
  • Linkmann, M
  • Marusic, I
  • Vinuesa, Ricardo
[EN] 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 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 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.; Understanding the role of coherent structures in the dynamics of turbulent flows is of high relevance for fluid dynamics, climate systems, and aerodynamics. The authors propose a deep learning approach to evaluate the importance of various types of coherent structure in the flow, to uncover main mechanisms of wall-bounded turbulence and develop techniques for its control., The deep-learning-model training was enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) at Berzelius (NSC), partially funded by the Swedish Research Council through grant agreement no. 2022-06725. This project has been partially funded by the Spanish Ministry of Science, Innovation, and University through the University Faculty Training (FPU) program with reference FPU19/02201 (AC). The data has been obtained with support of grant PID2021-128676OB-I00 funded by MCIN/AEI/10.13039/ 501100011033 and by "ERDF A way of making Europe", by the European Union (SH). RV acknowledges the financial support from ERC grant no. 2021-CoG-101043998, DEEPCONTROL. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. RD acknowledges the financial support from the Melbourne Postdoctoral Fellowship of the University of Melbourne. R.D., J.L., J.P.M., N.H., and I.M. are grateful to the ARC (Australian Research Council) for their continuous financial support.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

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

  • Cremades-Botella, Andrés|||0000-0002-7052-4913
  • Hoyas, Sergio|||0000-0002-8458-7288
  • Vinuesa, Ricardo
[EN] 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., The authors acknowledge Adrián Lozano-Durán and Álvaro MartíinezSánchez for their support with the validation of the causal nature of the SHAP values. The deep-learning-model training was enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) at Berzelius (NSC), partially funded by the Swedish Research Council through grant agreement no. 2022-06725. Part of the postprocessing was made in the supercomputer Sirius of the Universitat Politècnica de València. SH has the support of grant PID2021- 128676OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe , by the European Union. RV acknowledges the financial support from ERC grant no. 2021-CoG-101043998, DEEPCONTROL. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. Open access funding provided by Royal Institute of Technology
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Sequential and Parallel Algorithms to Compute Turbulent Coherent Structures

  • Gandía-Barberá, Sergio
  • Vinuesa, Ricardo
  • Cremades-Botella, Andrés|||0000-0002-7052-4913
  • Hoyas, Sergio|||0000-0002-8458-7288
  • Pérez Quiles, María Jezabel
[EN] The behavior of turbulent flows remains a significant unsolved problem in physics. Recently, a large quantity of effort has been directed toward understanding the non-linear interactions of the different flow structures in order to address this challenge. In this paper, different implementations of one exact method for identifying these structures are analyzed. This includes two sequential algorithms and a parallelizable one, developed to handle large-scale data efficiently. The new parallel algorithm offers significant advantages in handling the computational demands of large simulations, providing a more scalable solution for future research., S.H. and M.J.P.-Q. acknowledge the project PID2021-128676OB-I00 by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe", by the European Union (SH). R.V. acknowledges the financial support from ERC grant '2021-CoG-101043998, DEEPCONTROL'. The views and opinions expressed are, however, those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Deep Reinforcement Learning for Flow Control Exploits Different Physics for Increasing Reynolds Number Regimes

  • Varela-Martínez, Pau
  • García-Cuevas González, Luis Miguel|||0000-0001-9340-0617
  • Suárez Morales, Pol
  • Alcántara-Ávila, Francisco
  • Miró, Arnau
  • Rabault, Jean
  • Font, Bernat
  • Lehmkuhl, Oriol
  • Vinuesa, Ricardo
[EN] The increase in emissions associated with aviation requires deeper research into novel sensing and flow-control strategies to obtain improved aerodynamic performances. In this context, data-driven methods are suitable for exploring new approaches to control the flow and develop more efficient strategies. Deep artificial neural networks (ANNs) used together with reinforcement learning, i.e., deep reinforcement learning (DRL), are receiving more attention due to their capabilities of controlling complex problems in multiple areas. In particular, these techniques have been recently used to solve problems related to flow control. In this work, an ANN trained through a DRL agent, coupled with the numerical solver Alya, is used to perform active flow control. The Tensorforce library was used to apply DRL to the simulated flow. Two-dimensional simulations of the flow around a cylinder were conducted and an active control based on two jets located on the walls of the cylinder was considered. By gathering information from the flow surrounding the cylinder, the ANN agent is able to learn through proximal policy optimization (PPO) effective control strategies for the jets, leading to a significant drag reduction. Furthermore, the agent needs to account for the coupled effects of the friction- and pressure-drag components, as well as the interaction between the two boundary layers on both sides of the cylinder and the wake. In the present work, a Reynolds number range beyond those previously considered was studied and compared with results obtained using classical flow-control methods. Significantly different forms of nature in the control strategies were identified by the DRL as the Reynolds number Re increased. On the one hand, for Re¿1000, the classical control strategy based on an opposition control relative to the wake oscillation was obtained. On the other hand, for Re=2000, the new strategy consisted of energization of the boundary layers and the separation area, which modulated the flow separation and reduced the drag in a fashion similar to that of the drag crisis, through a high-frequency actuation. A cross-application of agents was performed for a flow at Re=2000, obtaining similar results in terms of the drag reduction with the agents trained at Re=1000 and 2000. The fact that two different strategies yielded the same performance made us question whether this Reynolds number regime (Re=2000) belongs to a transition towards a nature-different flow, which would only admits a high-frequency actuation strategy to obtain the drag reduction. At the same time, this finding allows for the application of ANNs trained at lower Reynolds numbers, but are comparable in nature, saving computational resources., The authors acknowledge the contribution of Maxence Deferrez to this work. R.V. acknowledges funding from the ERC through grant no. 2021-CoG-101043998, DEEPCONTROL
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Easy attention: A simple attention mechanism for temporal predictions with transformers

  • Sanchis-Agudo, Marcial
  • Wang, Yuning
  • Guastoni, Luca
  • Lim, Jasmin
  • Duraisamy, Karthik
  • Vinuesa, Ricardo
  • Arnau, Roger|||0000-0003-2544-8875
[EN] To improve the robustness of transformer neural networks used for temporal-dynamics prediction of chaotic systems, we propose a novel attention mechanism called easy attention, which we demonstrate in time-series reconstruction and prediction. While the standard self-attention only makes use of the inner product of queries and keys, it is demonstrated that the keys, queries, and softmax are not necessary for obtaining the attention score required to capture long-term dependencies in temporal sequences. Through the singular-value decomposition on the softmax attention score, we further observe that self-attention compresses the contributions from both queries and keys in the space spanned by the attention score. Therefore, our proposed easy-attention method directly treats the attention scores as learnable parameters. This approach produces excellent results when reconstructing and predicting the temporal dynamics of chaotic systems exhibiting more robustness and less complexity than self-attention or the widely used long short-term memory network. We show the improved performance of the easy-attention method in the Lorenz system, a turbulence shear flow, and a model of a nuclear reactor., R.V. acknowledges financial support from ERC Grant No. 21- CoG-101043998, DEEPCONTROL, and the EU Doctoral Network MODELAIR. Part of the ML-model testing was carried out using computational resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS). Jasmin Lim and Karthik Duraisamy are supported by ARPA-E under the project SAFARI: Secure Automation for Advanced Reactor Innovation at the University of Michigan. Roger Arnau was supported by a contract of PAID-01-21 from the Universitat Politècnica de València. The authors would like to acknowledge Professor Igor Meziæ for helpful feedback on the paper and Francisco-Javier Granados-Ortiz for fruitful discussions.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Tesis doctoral (DoctoralThesis)

Poiseuille and Couette flows: the effect of the small scales on the large structures

  • Gandía Barberá, Sergio
[ES] Diferentes flujos turbulentos en canales a diversos números de Reynolds de fricción son estudiados para elucidar los cambios en su organización de escalas turbulentas y la transferencia de energía originada por el esfuerzo cortante medio. Los flujos en canales representan el flujo más simple limitado por paredes, proporcionando un modelo simplificado de la dinámica de la turbulencia en geometrías mucho más complejas limitadas por paredes. En este estudio, nos centramos en flujos de Poiseuille, flujos de Couette y configuraciones combinadas. Por un lado, un flujo de canal de Poiseuille es un flujo impulsado por presión en el cual el fluido se mueve entre paredes paralelas debido a un gradiente de presión. Por otro lado, un flujo de canal de Couette es un flujo impulsado por esfuerzo cortante en el cual el fluido se mueve debido al movimiento relativo de una pared respecto a la otra, sin que intervenga un gradiente de presión. Las configuraciones combinadas contienen ambos mecanismos y cubren una transición escalonada desde un flujo puramente de Poiseuille hasta un flujo puramente de Couette. La física de ambos flujos se resuelve utilizando Simulación Numérica Directa (DNS) a números de Reynolds de fricción bajos a moderados, proporcionando información en todas las escalas turbulentas. El estudio analiza y caracteriza remolinos y estructuras coherentes, incluidas streaks y estructuras intensas de esfuerzos de Reynolds. Se calculan y comparan entre las configuraciones de flujo en canal parámetros geométricos, intensidad turbulenta, ubicación espacial y otros parámetros relevantes mediante gráficos probabilísticos y tendencias. En esta tesis, elucidamos que el esfuerzo cortante medio puede crear estructuras de gran escala promediadas en el tiempo, en forma de streaks y rolls, en configuraciones puramente de Couette y combinadas con suficiente contribución de esfuerzo cortante. Las estructuras de gran escala deben estar presentes para producir estadísticas relevantes del canal; por lo tanto, el tamaño mínimo de dominio en la dirección transversal está limitado por ellas. La generación de estas estructuras de gran escala está vinculada a un proceso autosostenido, el cual es sensible a fuerzas externas tales como la estratificación estable. De hecho, explicamos que una estratificación relativamente pequeña es capaz de romper este proceso autosostenido y, en consecuencia, prevenir el crecimiento de estructuras de gran escala. Además, estas estructuras de gran escala producen una reasignación de estructuras coherentes instantáneas a lo largo de la dirección transversal. Curiosamente, estas streaks y bursts instantáneas permanecen autosemejantes en cualquier configuración, independientemente de la presencia de estructuras de gran escala, de modo que sus proporciones dimensionales no difieren. Por último, al observar flujos de Couette térmicos pasivos, aparecen estructuras de gran escala promediadas en el tiempo en forma de regiones frías y calientes de temperatura., [CA] Diferents fluxos turbulents en canals a diversos números de Reynolds de fricció són estudiats per a elucidar els canvis en la seua organització d'escales turbulentes i la transferència d'energia originada per la cizalla mitjana. Els fluxos en canals representen el flux més simple limitat per parets, proporcionant un model simplificat de la dinàmica de la turbulència en geometries molt més complexes limitades per parets. En este estudi, ens centrem en fluxos de Poiseuille, fluxos de Couette i configuracions combinades. D'una banda, un flux de canal de Poiseuille és un flux impulsat per pressió en el qual el fluid es mou entre parets paral·leles a causa d'un gradient de pressió. D'altra banda, un flux de canal de Couette és un flux impulsat per cizalla en el qual el fluid es mou pel moviment relatiu d'una paret respecte a l'altra, sense que intervinga cap gradient de pressió. Les configuracions combinades contenen ambdós mecanismes i cobreixen una transició escalonada des d'un flux purament de Poiseuille fins a un flux purament de Couette. La física d'ambdós fluxos es resol utilitzant Simulació Numèrica Directa (DNS) a números de Reynolds de fricció baixos a moderats, proporcionant informació en totes les escales turbulentes. L'estudi analitza i caracteritza remolins i estructures coherents, incloent-hi streaks i estructures intenses d'esforços de Reynolds. Es calculen i comparen entre les configuracions de flux en canal paràmetres geomètrics, intensitat turbulenta, ubicació espacial i altres paràmetres rellevants mitjançant gràfics probabilístics i tendències. En esta tesi, elucidarem que la cizalla mitjana pot crear estructures de gran escala promediades en el temps, en forma de streaks i rolls, en configuracions purament de Couette i combinades amb suficient contribució de cizalla. Les estructures de gran escala han d'estar presents per a produir estadístiques rellevants del canal; per tant, la mida mínima del domini en la direcció transversal queda limitada per elles. La generació d'estes estructures de gran escala està vinculada a un procés autosostingut, el qual és sensible a forces externes com ara l'estratificació estable. De fet, expliquem que una estratificació relativament xicoteta és capaç de trencar este procés autosostingut i, en conseqüència, prevenir el creixement d'estructures de gran escala. A més, estes estructures de gran escala produeixen una reallocació d'estructures coherents instantànies al llarg de la direcció transversal. Curiosament, estes streaks i bursts instantanis romanen autosemejants en qualsevol configuració, independentment de la presència d'estructures de gran escala, de manera que les seues proporcions dimensionals no diferixen. Finalment, en observar fluxos de Couette tèrmics passius, apareixen estructures de gran escala promediades en el temps en forma de regions fredes i calentes de temperatura., [EN] Different turbulent channel flows at diverse friction Reynolds numbers are studied to elucidate the changes in their turbulent scale organisation and energy transfer originating from the mean shear. Channel flows represent the simplest wall-bounded flow, providing a simplified model of the turbulence dynamics in much more complex wall-bounded geometries. In this study, we focus on Poiseuille flows, Couette flows and combined configurations. On the one hand, a Poiseuille channel flow is a pressure-driven flow in which the fluid moves between parallel walls due to a pressure gradient. On the other hand, a Couette channel flow is a shear-driven flow in which the fluid moves due to the relative movement of one wall to the other with no pressure gradient involved. The combined configurations contain both mechanisms and cover a stepped transition from a pure Poiseuille flow to a pure Couette flow. The physics of both flows is resolved using Direct Numerical Simulation (DNS) at low to moderate Reynolds friction numbers, providing information at all turbulent scales. The study analyses and characterises eddies and coherent structures, including streaks and intense Reynolds stress structures. Geometrical parameters, turbulent intensity, spatial location and other relevant parameters are calculated and compared among channel flow configurations using probabilistic plots and trends. In this thesis, we elucidate that the mean shear can create time-averaged largescale structures, in the form of streaks and rolls, in pure Couette and combined configurations with enough shear contribution. Large-scale structures must be present in order to produce relevant channel statistics; therefore, the minimum domain spanwise size is limited by them. The generation of these large-scale structures is linked to a self-sustaining process, which is sensitive to external forces such as stable stratification. Indeed, we explain that relatively small stratification is able to break this self-sustaining process and consequently prevents the growth of large-scale structures. Furthermore, these large-scale structures produce a reallocation of instantaneous coherent structures along the spanwise direction. Curiously enough, these instantaneous streaks and bursts remain self-similar in any configuration, independently of the presence of large-scale structures, so their dimensional ratios do not differ. Last but not least, when observing passive thermal Couette flows, similar time-averaged large-scale structures appear in the form of cold and hot temperature regions., The present study was developed at the Instituto Universitario de Investigación de Matemática Pura y Aplicada (IUMPA) of the Universitat Politècnica de València (UPV). This work was supported by MINECO/FEDER under grants RTI2018-102256-B-I00 and ENE2015-71333-R, and by MCIN/AEI under grant PID2021-128676OB-I00, co-funded by the European Union through the European Regional Development Fund (ERDF) “A Way of Making Europe”. The computations were made possible thanks to generous grants of computing time from the Barcelona Supercomputing Centre (references FI-2018-1-0037 and AECT-2020-2-0005) and from the supercomputing centre of the Universitat Politècnica de València. The work of M. Oberlack was supported by the German Research Foundation (DFG) under grant OB96/39-1. FAA is partially funded by the GVA/FEDER project ACIF2018. S.H. and M.J.P.Q. acknowledge support from project PID2021-128676OB-I00 by MCIN/AEI/10.13039/501100011033 and ERDF “A Way of Making Europe”. R.V. acknowledges financial support from the ERC grant 2021-CoG-101043998 (DEEPCONTROL).
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

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

  • Nagib, Hassan
  • Vinuesa, Ricardo
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] 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-tau < 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 nine out of 15 cases of DNS data we examined where Re-tau < 2000, the NSP did not contain either of the proposed trends. As Re-tau exceeds around 2000 a 1/4-power, reflecting the "bounded-dissipation" predictions of Chen and Sreenivasan ["Law of bounded dissipation and its consequences in turbulent wall flows," J. Fluid Mech. 933, A20 (2022); "Reynolds number asymptotics of wall-turbulence fluctuations," J. Fluid Mech. 976, A21 (2023)] and data analysis of Monkewitz ["Reynolds number scaling and inner-outer overlap of stream-wise Reynoldss stress in wall turbulence," arXiv:2307.00612 (2023)], develops near y(+) = 1000 and expands with Reynolds numbers extending to 1000 < y(+) < 10 000 for Re-tau 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 1000 by Marusic et al. ["Attached eddy model of wall turbulence," Annu. Rev. Fluid Mech. 51, 49-74 (2019); "The logarithmic variance of streamwise velocity and conundrum in wall turbulence," J. Fluid Mech. 933, A8 (2022)] according to a "wall-scaled eddy model." (c) 2024 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/)., SH was funded by project No. PID2021-128676OB-I00 by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe," by the European Union (SH). RV acknowledges the financial support from ERC Grant No. 2021-CoG-101043998, DEEPCONTROL.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Separation control applied to the turbulent flow around a NACA4412 wing section

  • Wang, Yuning
  • Mallor, Fermín
  • Mariani, Raffaello
  • Vinuesa, Ricardo
  • Guardiola, Carlos|||0000-0002-3150-8566
[EN] We carried out high-resolution large-eddy simulations (LESs) to investigate the effects of several separation-control approaches on a NACA4412 wing section with spanwise width of Lz = 0.6 at an angle of attack of AoA = 11 degrees at a Reynolds number of Rec = 200,000 based on chord length c and free-stream velocity U infinity. Two control strategies were considered: (1) steady uniform blowing and/or suction on the suction and/or pressure sides, and (2) periodic control on the suction side. A wide range of control configurations were evaluated in terms of aerodynamic efficiency (i.e., lift-to-drag ratio) and separation delay. Uniform blowing and/or suction effectively delayed flow separation, leading to a lift increase of up to 11%, but yielded only marginal improvements in aerodynamic efficiency. In contrast, periodic control neither enhanced separation delay nor improved efficiency. A detailed analysis of the interaction between uniform blowing and/or suction and turbulent boundary layers (TBLs) over the wing was performed, including assessments of (1) integral boundary-layer quantities, (2) turbulence statistics, and (3) power-spectral densities. Significant modifications in Reynolds stresses and spectral characteristics were observed. To the authors' best knowledge, this is the first numerical study utilizing high-resolution LESs to provide comprehensive assessments on separation control., Numerical studies were enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) at PDC center for High Performance Computing in KTH (Stockholm). RV acknowledges the financial support from ERC grant no. '2021-CoG-101043998, DEEPCONTROL'. Views and opinions expressed are those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Inferring wall-bounded coherent structures from two-dimensional turbulent fields via SHAP analysis

  • Molina-Casino, Samuel
  • Cremades, Andrés
  • Cardesa, Jose I.
  • Chedevergne, François
  • Vinuesa, Ricardo
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] The analysis of coherent structures in turbulent flows is essential for understanding and controlling turbulence. Explainable Deep Learning (XDL) has been successfully used to identify the most important flow structures in turbulent channels (Cremades et al., 2024), but its computational cost grows significantly as Reynolds number increases. This study adapts the XDL framework to assess whether observations from time-resolved volumetric data of direct numerical simulations (DNS) can be inferred using two-dimensional (2D) slices. Limiting the analysis to 2D subsets reduces the cost of three-dimensional (3D) post-processing and facilitates its application to planar particle image velocimetry (PIV) experimental data. We employ SHapley Additive exPlanations (SHAP), a game-theoretic method, to identify dynamically relevant flow regions by assigning importance scores. Using DNS data from a turbulent channel flow at a friction Reynolds number of 125, we compare SHAP-based structures derived from 3D and 2D data with classically studied structures, including Q events, streaks, and vortex clusters. Our findings show that SHAP-based structures from 2D data statistically agree with those from full 3D fields and previously reported in the literature. These results underscore the potential of SHAP analyses for studying coherent structures with 2D data, enabling experimental and DNS research at higher Reynolds number., The data has been obtained with support of grant PID2021-128676OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe", by the European Union (SH) . RV acknowledges the financial support from ERC grant no. 2021-CoG-101043998, DEEPCONTROL. Views and opinions expressed are however those of the author (s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Deep-learning-based assessment of skin friction in wall-bounded turbulence

  • Hoyas, Sergio|||0000-0002-8458-7288
  • Benedikt, Nils
  • Cremades, Andrés
  • Vinuesa, Ricardo
[EN] This work investigates the influence of classically coherent structures on wall-shear stress and energy dissipation in turbulent channel flow, utilizing direct numerical simulations (DNS) data and explainable deep learning (XDL). Sweeps, defined as regions of low streamwise velocity moving toward the wall, are found to be the most influential structures for both energy dissipation and drag. Moreover, the volume of these key structures falls within a narrow range, making it easier to identify the most significant ones. Consequently, this work paves the way for the development of novel, highly efficient turbulence-control strategies for the reduction of drag., S.H. was funded by Project No. PID2021-128676OB-I00 by Ministerio de Ciencia e Innovacion, MCIN/AEI/10.13039/203501100011033 and by "ERDF A way of making Europe," by the European Union. R.V. acknowledges the financial support from ERC Grant No. "2021-CoG-101043998, DEEPCONTROL." The views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Three-dimensional ESRGAN for super-resolution reconstruction of turbulent flows with tricubic interpolation-based transfer learning

  • Yu, Linqui
  • Yousif, Mustafa Z.
  • Zhang, Meng
  • Vinuesa, Ricardo
  • Lim, Hee-Chang
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] Turbulence is a complicated phenomenon because of its chaotic behavior with multiple spatiotemporal scales. Turbulence also has irregularity and diffusivity, making predicting and reconstructing turbulence more challenging. This study proposes a deep-learning approach to reconstruct three-dimensional (3D) high-resolution turbulent flows from spatially limited data using a 3D enhanced super-resolution generative adversarial networks (3D-ESRGAN). In addition, a novel transfer-learning method based on tricubic interpolation is employed. Turbulent channel flow data at friction Reynolds numbers R e tau = 180 and R e tau = 500 were generated by direct numerical simulation (DNS) and used to estimate the performance of the deep-learning model as well as that of tricubic interpolation-based transfer learning. The results, including instantaneous velocity fields and turbulence statistics, show that the reconstructed high-resolution data agree well with the reference DNS data. The findings also indicate that the proposed 3D-ESRGAN can reconstruct 3D high-resolution turbulent flows even with limited training data., This work was supported by "Human Resources Program in Energy Technology" of the Korea Institute of Energy Technology Evaluation and Planning (KETEP), granted financial resource from the Ministry of Trade, Industry & Energy, Republic of Korea (No. 20214000000140). In addition, this work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. 2019R1I1A3A01058576). This work was also supported by the National Supercomputing Center with supercomputing resources including technical support (No. KSC-2022-CRE-0282). R.V. acknowledges the financial support from the ERC Grant No. 2021-CoG-101043998, DEEPCONTROL. S.H. was funded by Contract No. PID2021-128676OB-I00 of Ministerio de Ciencia, innovacion y Universidades/FEDER.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

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

  • Yousif, Mustafa Z.
  • Yu, Linqui
  • Vinuesa, Ricardo
  • Lim, Hee-Chang
  • Hoyas, Sergio|||0000-0002-8458-7288
[EN] 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 in the complexity of the experimental setup and the storage cost can be achieved., This work was supported by Human Resources Program in Energy Technology' of the Korea Institute of Energy Technology Evaluation and Planning (KETEP), granted financial resource from the Ministry of Trade, Industry & Energy, Republic of Korea (no. 20214000000140). In addition, this work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (no. 2019R1I1A3A01058576). This work was also supported by the National Supercomputing Center with supercomputing resources including technical support (KSC-2021-CRE-0244). R.V. acknowledges the financial support from the ERC Grant No. "2021-CoG-101043998, DEEPCONTROL". SH was funded by Contract No. PID2021-128676OB-I00 of Ministerio de Ciencia, innovacion y Universidades/ FEDER.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

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

  • Cremades-Botella, Andrés|||0000-0002-7052-4913
  • Hoyas, Sergio|||0000-0002-8458-7288
  • Vinuesa, Ricardo
[EN] 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 that explainability techniques, and in particular additive- feature-attribution methods, are crucial for implementing interpretable and physics-compliant deep-learning models in the fluid-mechanics field., RV acknowledges the financial support from ERC grant no. 2021-CoG-101043998, DEEPCONTROL. Views and opinions expressed are however those of the author (s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. SH is grateful for the grant PID2021-128676OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A Way of Making Europe", by the European Union
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Decoding complexity through machine learning is redefining scientific discovery

  • Vinuesa, Ricardo
  • Cinnella, Paola
  • Rabault, Jean
  • Azizpour, Hossein
  • Bauer, Stefan
  • Brunton, Bingni W.
  • Elofsson, Arne
  • Jarlebring, Elias
  • Kjellström, Hedvig
  • Markidis, Stefano
  • Marlevi, David
  • Garcia-Martinez, Javier
  • Brunton, Steven L.
  • Garcia-Martinez, Javier|||/items/421c2466-c63a-45d7-b799-c59d51faf1bc
As scientific instruments and the literature generate ever larger volumes of data, machine learning (ML) has become essential for organizing, analyzing and interpreting complex information. This Perspective examines how ML accelerates discovery across disciplines, with examples such as brain mapping and exoplanet detection. It also considers situations with different levels of prior knowledge about the underlying phenomenon, outlining strategies to address limitations and exploit ML effectively. Although growing reliance on ML raises challenges for research practice and validation, it is reshaping scientific methods and expanding what can be studied. We also highlight foundation models as a promising route to faster, broader scientific discovery.