Memory-Efficient Adjoints via Graph Partitioning

Indoor environmental quality is crucial for human health and comfort, necessitating precise and efficient computational methods to optimise indoor climate parameters. Recent advancements in machine learning (ML) and computational fluid dynamics (CFD) are promising. However, applying ML to complex building airflow presents challenges. This research aims to investigate the integration of ML with CFD in the context of built environment applications using a systematic review approach. It highlights a critical knowledge gap: the need to synthesise innovative approaches that address the limitations of indoor modelling using data-driven ML methods. The review examines contemporary literature, identifying current developments and suggesting potential future directions. It delves into the innovations in combining ML with CFD to predict thermal comfort and indoor air quality, uncovering key limitations such as the lack of high-quality experimental data for training and validation, the computational complexity of detailed CFD simulations, and the interpretability issues of ‘black-box’ ML models. The emergence of data-driven techniques in fluid mechanics offers promising prospects for modelling in the built environment. Future research should focus on incorporating physics-based rules in ML models, adapting turbulence closure models for indoor flows, and enhancing model validation using real-world datasets. The research emphasises the synergistic relationship between ML and CFD; it proposes pathways to overcome current limitations, aiming to enhance the precision and efficiency of indoor environment modelling through their integration.

Computational Fluid Dynamics (CFD) is regarded as an important tool for analyzing the flow field, thermal environment, and air quality around the built environment. However, for built environment applications, the high computational cost of CFD hinders large-scale scenario simulation and efficient design optimization. In the field of built environment research, surrogate modeling has become a key technology to connect the needs of high-fidelity CFD simulation and rapid prediction, whereas the low-dimensional nature of traditional surrogate models is unable to match the physical complexity and prediction needs of built flow fields. Therefore, combining machine learning (ML) with CFD to predict flow fields in built environments offers a promising way to increase simulation speed while maintaining reasonable accuracy. This review briefly reviews traditional surrogate models and focuses on ML-based surrogate models, especially the specific application of neural network architectures in rapidly predicting flow fields in the built environment. The review indicates that ML accelerates the three core aspects of CFD, namely mesh preprocessing, numerical solving, and post-processing visualization, in order to achieve efficient coupled CFD simulation. Although ML surrogate models still face challenges such as data availability, multi-physics field coupling, and generalization capability, the emergence of physical information-driven data enhancement techniques effectively alleviates the above problems. Meanwhile, the integration of traditional methods with ML can further enhance the comprehensive performance of surrogate models. Notably, the online ministry of trained ML models using transfer learning strategies deserves further research. These advances will provide an important basis for advancing efficient and accurate operational solutions in sustainable building design and operation.

RoseNNa: A performant, portable library for neural network inference with application to computational fluid dynamics

An investigation of the thermal behavior of constructal theory-based pore-scale porous media by using a combination of computational fluid dynamics and machine learning

In hydraulic engineering, accurately predicting energy dissipation at vertical drop structures is crucial for ensuring structural integrity and optimizing design efficiency. This study presents a comparative analysis of energy dissipation calculations using two methodologies: numerical modeling through Computational Fluid Dynamics (CFD) and advanced Machine Learning (ML) techniques. The study is performed at vertical drop structure at Lower Gogera Branch Canal (LGBC) in Pakistan, experimental data has been utilized alongside CFD simulations to establish a reliable benchmark for energy dissipation at a vertical drop of 3.69 meters, with a discharge rate of 63.73 m³/s. The CFD method employs the Volume of Fluid (VOF) technique to simulate flow dynamics, providing precise insights into energy dissipation. In parallel, we implement various ML algorithms, including decision trees, random forests, and neural networks, to model energy dissipation based on key input parameters such as flow rate, height of fall, and downstream conditions. The performance of the ML models is evaluated against the numerical results, allowing for a comprehensive comparison of their predictive capabilities. Initial findings indicate that while the CFD model shows a high level of accuracy with discrepancies within 5% of empirical measurements, the ML techniques also demonstrate significant potential in capturing complex, nonlinear relationships inherent in energy dissipation phenomena. The use of ML not only provides a faster alternative for predictions but also offers insights into optimizing design parameters in hydraulic structures. This research highlights the transformative impact of AI methodologies on hydraulic engineering, illustrating how integrating ML can enhance traditional modeling techniques. The results contribute to the ongoing discourse on the application of artificial intelligence in water resources management, paving the way for more efficient and sustainable engineering practices.

In several problems involving fluid flows, computational fluid dynamics (CFD) provides detailed quantitative information and allows the designer to successfully optimize the system by minimizing a cost function. Sometimes, however, one cannot improve the system with CFD alone, because a suitable cost function is not readily available; one notable example is diagnosis in medicine. The application considered here belongs to the field of rhinology; a correct air flow is key for the functioning of the human nose, yet the notion of a functionally normal nose is not available and a cost function cannot be written. An alternative and attractive pathway to diagnosis and surgery planning is offered by data-driven methods. In this work, we consider the machine learning study of nasal impairment caused by anatomic malformations, with the aim of understanding whether fluid dynamic features, available after a CFD analysis, are more effective than purely geometric features at the training of a neural network for regression. Our experiments are carried out on an extremely simplified anatomic model and a correspondingly simple CFD approach; nevertheless, they show that flow-based features perform better than geometry-based ones and allow the training of a neural network with fewer inputs, a crucial advantage in fields like medicine.

Optimization methods have been widely applied to the aerodynamic design of gas turbine blades. While applying optimization to high-fidelity computational fluid dynamics (CFD) simulations has proven capable of improving engineering design performance, a challenge has been overcoming the prolonged run-time due to the computationally expensive CFD runs. Reduced-order models and, more recently, machine learning methods have been increasingly used in gas turbine studies to predict performance metrics and operational characteristics, model turbulence, and optimize designs. The application of machine learning methods allows for utilizing existing knowledge and datasets from different sources, such as previous experiments, CFD, low-fidelity simulations, 1D or system-level studies. The present study investigates inserting a machine learning model that utilizes such data into a high-fidelity CFD driven optimization process, and hence effectively reduces the number of required evaluations of the CFD model. Artificial Neural Network (ANN) models were trained on data from over three thousand two-dimensional (2D) CFD analyses of turbine blade cross-sections. The trained ANN models were then used as surrogates in a nested optimization process alongside a full three-dimensional Navier–Stokes CFD simulation. The much lower evaluation cost of the ANN model allows for tens of thousands of design evaluations to guide the search of the best blade profiles to be used in the more expensive, high-fidelity CFD runs, improving the progress of the optimization while reducing the required computation time. It is estimated that the current workflow achieves a five-fold reduction in computational time in comparison to an optimization process that is based on three-dimensional (3D) CFD simulations alone. The methodology is demonstrated on the NASA/General Electric Energy Efficient Engine (E3) high pressure turbine blade and found Pareto front designs with improved blade efficiency and power over the baseline. Quantitative analysis of the optimization data reveals that some design parameters in the present study are more influential than others, such as the lean angle and tip scaling factor. Examining the optimized designs also provides insight into the physics, showing that the optimized designs have a lower amount of pressure drop near the trailing edge, but have an earlier onset of pressure drop on the suction side surface when compared to the baseline design, contributing to the observed improvements in efficiency and power.

Numerical investigation of water forced convection inside a copper metal foam tube: Genetic algorithm (GA) based fuzzy inference system (GAFIS) contribution with CFD modeling

This thesis investigates the integration of machine learning techniques into computational fluid dynamics (CFD) to improve simulation accuracy and computational performance. It proposes the use of the Gaussian Mixture Model (GMM) and unsupervised clustering as novel tools for mesh adaptation and flow feature detection. The methodology applies GMM to segment the computational domain into regions with distinct physical behaviour, viscous/turbulent and inviscid/irrotational. This clustering-based p-adaptation approach removes the need for predefined thresholds, enabling a fully automated and localised adjustment of the solution's polynomial degree. As a result, mesh refinement becomes more efficient and precise. A series of numerical experiments, including simulations of flow past a cylinder at varying Reynolds numbers, demonstrate the ability of the method to balance computational cost and solution fidelity. The framework is also tested on real-world engineering cases, such as wind turbine wake simulations, confirming its scalability and robustness. The results show substantial gains in efficiency and adaptability, establishing this approach as a promising solution for a wide range of CFD applications. By embedding machine learning into high-order solvers, this work advances the state-of-the-art and opens new directions for adaptive simulation strategies. It also lays the groundwork for future research in error-driven refinement, 3D turbulence modeling, and the broader application of AI in fluid mechanics. Keywords: Computational fluid dynamics, machine learning, Gaussian Mixture Model, p-adaptation, flow feature detection, mesh refinement, high-order methods. RESUMEN Esta tesis examina la integración de técnicas de aprendizaje automático en la dinámica de fluidos computacional (CFD) para mejorar la precisión de las simulaciones y reducir el coste computacional. Se propone el uso del Modelo de Mezclas Gaussianas (GMM) y métodos de agrupamiento no supervisado para desarrollar nuevas estrategias de adaptación de malla y detección de regiones del flujo sin recurrir a umbrales predefinidos. La metodología aplica GMM para segmentar automáticamente el dominio en regiones con diferentes características físicas, como zonas viscosas/turbulentas y no viscosas/irrotacionales. Esta clasificación guía un proceso de p-adaptación, ajustando el orden polinomial de la solución numérica de forma localizada, lo que permite un refinamiento automático, preciso y eficiente. Diversos experimentos numéricos, incluyendo simulaciones del flujo alrededor de un cilindro para distintos números de Reynolds, demuestran que el enfoque propuesto equilibra de manera eficaz el coste computacional y la calidad de la solución. Además, su aplicabilidad se valida en casos reales como las estelas de aerogeneradores, evidenciando la escalabilidad del método. Los resultados confirman que estas técnicas mejoran la eficiencia y adaptabilidad de los solucionadores de alto orden, lo que las convierte en herramientas prometedoras para aplicaciones complejas en CFD. La tesis contribuye así al estado del arte al integrar aprendizaje automático con métodos numéricos avanzados, y sienta las bases para investigaciones futuras sobre refinamiento adaptativo basado en errores, simulaciones tridimensionales de turbulencia y uso extendido de IA en mecánica de fluidos. Palabras clave: dinámica de fluidos computacional, aprendizaje automático, Modelo de Mezclas Gaussianas, p-adaptación, detección de características, refinamiento de mallas, métodos de alto orden. RÉSUMÉ Cette thèse explore l’intégration des techniques d’apprentissage automatique dans la dynamique des fluides numérique (CFD) afin d’améliorer la précision des simulations et la surcharge de calcul. En utilisant le modèle de mélanges gaussiens (GMM) et des méthodes de regroupement non supervisé, cette recherche introduit des approches innovantes pour l’adaptation du maillage et la détection des structures d’écoulement. La méthodologie se concentre sur l’application du GMM pour partitionner les champs d’écoulement en régions distinctes, régions distinctes, différenciant les zones visqueuses/turbulentes des zones inviscides/irrotationnelles. Cette technique de p-adaptation basée sur le regroupement élimine le besoin de seuils prédéfinis, permettant ainsi un raffinement du maillage automatisé, efficace et précis. De nombreuses expériences numériques sur divers cas tests, notamment l’écoulement autour d’un cylindre à différents nombres de Reynolds, démontrent la capacité de la méthode proposée à équilibrer le coût computationnel et la fidélité de la solution. Des applications en ingénierie, telles que la simulation du sillage d’une éolienne, illustrent davantage l’évolutivité des méthodes proposées. Les résultats montrent des améliorations significatives en termes du coût computationnel et d’adaptabilité, rendant ces techniques particulièrement adaptées à un large éventail d’applications en CFD. Ce travail contribue à l’état de l’art en intégrant de manière fluide l’apprentissage automatique dans les solveurs CFD d’ordre élevé, ouvrant ainsi la voie à de futures avancées dans les méthodes de simulation adaptative. Les méthodologies proposées établissent une base pour l’exploration du raffinement basé sur l’erreur, la modélisation tridimensionnelle de la turbulence et l’adoption plus large de l’apprentissage automatique dans la recherche en dynamique des fluides.

The development of near-term quantum computers, referred to as Noisy Intermediate-Scale Quantum (NISQ) computers, has progressed rapidly in the past few years resulting in several quantum computers which vary in their underlying technology and physical constraints. The performance of these computers also varies from one quantum algorithm to another. To enable efficient selection of the quantum computer that provides the highest output fidelity for a given application, an accurate noise modeling of each quantum hardware is required. However, noise modeling for a given application is a complex problem because of the unknown interaction between the quantum circuit parameters and the noise parameters of NISQ devices. We propose the use of Machine Learning (ML) to model the performance of different quantum computers at the application level. The ML models predict the output fidelity of the quantum application executed on different quantum computers given their publicly available physical constraints. We use a diverse training dataset to cover the key features for application-level benchmarking of the quantum hardware. Our results obtained from different superconducting quantum devices show that our proposed ML models enable cost-effective quantum computer selection for different quantum applications with different fidelity metrics.

Over the years, computational fluid dynamics (CFD) has undergone significant evolution, establishing it as an essential tool for simulating complex fluid flow phenomena. The integration of deep learning (DL) and machine learning (ML) methodologies with CFD has recently emerged as a promising approach to enhance the accuracy, efficiency, and automation of simulations. This comprehensive review article explores the potential of ML and DL in advancing CFD, providing valuable insights into their transformative impact on the field. In this study, the large eddy simulation (LES), reduced-order models (ROMs), Reynolds-averaged Navier-Stokes (RANS), and direct numerical simulation (DNS) techniques, along with the role of ML in improving these models within CFD, are thoroughly discussed. It was demonstrated that ML and DL techniques can accelerate high-fidelity simulations, provide turbulence models with varying levels of precision, and generate ROMs that surpass the accuracy achieved through conventional methods. Additionally, this paper presents a broad perspective, highlighting recent advancements, opportunities, and unresolved challenges in the field.

In the field of energy applied to fluid dynamics, the net power output of a system is dependant upon the thermal and hydraulic behaviour of the flow in interaction with surrounding objects. For cooling systems applications, the interest relies on the understanding and optimisation of the Thermo-Hydraulic Performance (THP) research. The THP optimisation relevant to this research is that of shape parametrisation. It has been demonstrated that utilising a biomimetic geometry on the fluid-facing surface of a pipe, based on the skin of fast sharks, improves hydraulic performance. However, the design of such geometry introduces a number of challenges that need to be addressed. First, the thermal and hydraulic mechanisms of the flow have a coupled nature, with a trade-off between them leading to a Multi-Objective Optimisation (MOO) problem. Second, the use of Computational Fluid Dynamics (CFD)to analyse the coupled flow behaviour. This is a cost-effective alternative to experimental testing and a favourable tool for geometry parametrisation. However, CFD implies a back-and-forth process between design and modelling over several iterations of shape parameters, which can make the development of optimal geometries highly time-consuming for the required level of accuracy. Therefore, industry demands for increasing levels of accuracy and the need for optimising the design process itself prompts the search for new means to undertake these tasks. In the ambition to address the above challenges within the computational design process, this thesis puts forward a comparative data-driven computational framework with an associated open-source software. The software generates both a database containing CFD high-fidelity models of parametrised geometries, as well as surrogate models that are trained with such data to approximate the THP optimisation process. The thesis puts forward three novelties. First, the study of the influence that simplified biomimetic geometries have in the coupled thermo-hydraulic behaviour of the flow in pipes. Second, a comparative analysis using alternative Machine Learning (ML) techniques in order to identify the advantages and drawbacks of each algorithm across different approaches. The considered ML methods are Radial Basis Function (RBF) interpolation, Neural Networks (NN) and Gaussian Processes (GP),with the addition of Proper Orthogonal Decomposition (POD) as a dimensional reduction method prior to the models’ training to potentially ease the computational expense. Third, the framework’s main direction is to use ML algorithms to enrich the search for an optimal geometry within the considered parametric space. The efforts made to reach the objectives of this research have resulted into the open-source software Hammerhead integrated environment for high-fidelity and ML based computational design optimisation. This new framework is developed and implemented using Python3 language, including compatibility with OpenFOAM open-source CFD platform.

Couplings connect the spinning shafts of driving and driven machines in the industry. A coupling guard encloses the coupling to protect personnel from the high-speed rotating coupling. The American Petroleum Institute API publishes standards that restrict the overheating of the coupling guards due to windage caused by the spinning shaft. Based on the most recent version of API 671, the peak temperature for the coupling guard should not exceed 60 °C. This paper proposes a machine learning (ML) model and an empirical formula to predict the maximum guard temperature and power loss. The ML models use a database obtained from simulated computational fluid dynamics (CFD) cases for different coupling guards under various conditions. Also, the paper provides validation for the CFD models with experimental tests for different cases. The proposed ML model uses eight different input parameters to predict temperature and power loss. The model shows an accurate prediction for a varied number of CFD cases. The performance of the generated model has been verified with the experimental results. Also, an empirical formula has been created using the same database from CFD results. The results show that the ML model has better prediction accuracy than the empirical formula for predicting peak temperature and power loss for all cases.

Purpose This study aims to introduce a novel machine learning feature vector (MLFV) method to bring machine learning to overcome the time-consuming computational fluid dynamics (CFD) simulations for rapidly predicting turbulent flow characteristics with acceptable accuracy. Design/methodology/approach In this method, CFD snapshots are encoded in a tensor as the input training data. Then, the MLFV learns the relationship between data with a rod filter, which is named feature vector, to learn features by defining functions on it. To demonstrate the accuracy of the MLFV, this method is used to predict the velocity, temperature and turbulent kinetic energy fields of turbulent flow passing over an innovative nature-inspired Dolphin turbulator based on only ten CFD data. Findings The results indicate that MLFV and CFD contours alongside scatter plots have a good agreement between predicted and solved data with R2 ≃ 1. Also, the error percentage contours and histograms reveal the high precisions of predictions with MAPE = 7.90E-02, 1.45E-02, 7.32E-02 and NRMSE = 1.30E-04, 1.61E-03, 4.54E-05 for prediction velocity, temperature, turbulent kinetic energy fields at Re = 20,000, respectively. Practical implications The method can have state-of-the-art applications in a wide range of CFD simulations with the ability to train based on small data, which is practical and logical regarding the number of required tests. Originality/value The paper introduces a novel, innovative and super-fast method named MLFV to address the time-consuming challenges associated with the traditional CFD approach to predict the physics of turbulent heat and fluid flow in real time with the superiority of training based on small data with acceptable accuracy.

Computational fluid dynamics (CFD) simulating is a useful methodology for reduction of experiments and their associated costs. Although the CFD could predict all hydro-thermal parameters of fluid flows, the connections between such parameters with each other are impossible using this approach. Machine learning by the artificial intelligence (AI) algorithm has already shown the ability to intelligently record engineering data. However, there are no studies available to deeply investigate the implicit connections between the variables resulted from the CFD. The present investigation tries to conduct cooperation between the mechanistic CFD and the artificial algorithm. The genetic algorithm is combined with the fuzzy interface system (GAFIS). Turbulent forced convection of Al2O3/water nanofluid in a heated tube is simulated for inlet temperatures (i.e., 305, 310, 315, and 320 K). GAFIS learns nodes coordinates of the fluid, the inlet temperatures, and turbulent kinetic energy (TKE) as inputs. The fluid temperature is learned as output. The number of inputs, population size, and the component are checked for the best intelligence. Finally, at the best intelligence, a formula is developed to make a relationship between the output (i.e. nanofluid temperatures) and inputs (the coordinates of the nodes of the nanofluid, inlet temperature, and TKE). The results revealed that the GAFIS intelligence reaches the highest level when the input number, the population size, and the exponent are 5, 30, and 3, respectively. Adding the turbulent kinetic energy as the fifth input, the regression value increases from 0.95 to 0.98. This means that by considering the turbulent kinetic energy the GAFIS reaches a higher level of intelligence by distinguishing the more difference between the learned data. The CFD and GAFIS predicted the same values of the nanofluid temperature.