Description:FIELD OF DISCLOSURE
[0001] The present disclosure relates generally relates to the field of computational fluid dynamics (CFD) and artificial intelligence-based modeling. More specifically, it pertains to a hybrid CFD-ANN framework system for real-time fluid flow analysis.
BACKGROUND OF THE DISCLOSURE
[0002] Computational fluid dynamics (CFD) has long served as a cornerstone in the simulation and analysis of fluid behavior, rooted in solving the robust and highly nonlinear Navier–Stokes equations. These governing equations embody conservation of mass, momentum, and energy, and form the mathematical underpinning of fluid mechanics. Early computational efforts, such as those by Lewis Fry Richardson, envisioned discretized computations over physical domains foreshadowing modern CFD even though they struggled with primitive computing capabilities. The progression of hardware, typified by ENIAC-era simulations and pioneering work at laboratories like Los Alamos, ultimately propelled efforts to tackle more complex, transient flows through methods like vorticity-stream function formulations and particle-in-cell techniques.
[0003] Traditional CFD techniques typically employ finite difference, finite volume, or finite element schemes that discretize space (and time) to numerically solve PDEs. Among these, the hybrid difference schemes blending central difference (quadratic accuracy but prone to stability issues at high convective dominance) with upwind differencing (lower accuracy, but superior stability under strong convection) epitomize early attempts to balance performance and realism in computational treatment of convection-diffusion problems.
[0004] Yet, even with modern HPC infrastructure, CFD remains resource-intensive especially for three-dimensional, unsteady, or turbulent flows where fine spatial-temporal resolution is essential for capturing subtle dynamics. The computational burden grows exponentially with domain complexity, physical fidelity, and temporal duration, limiting the feasibility of real-time or interactive fluid analysis.
[0005] Driven by growing availability of high-fidelity simulation data and advances in machine learning (ML), researchers have increasingly sought data-driven approaches to accelerate or enhance CFD. Among these, surrogate modeling emerges as a particularly attractive strategy: ML models such as response surface methods, Gaussian processes, or artificial neural networks (ANNs) can learn the functional mapping between input parameters and fluid responses, delivering rapid inference once trained.
[0006] ANNs, in particular, excel at approximating nonlinear, high-dimensional relationships and have been applied in diverse fluid-related scenarios. For example, ANFIS models (Combining neural networks with fuzzy logic) have been used to predict temperature and velocity in solar chimneys with R² values above 0.97, highlighting their predictive power and efficiency compared to CFD alone.
[0007] Other endeavors include CFD-ANN hybrid schemes for heat exchangers (specifically tube banks), where CFD-generated data trained ANNs to predict thermal-hydraulic parameters like friction factors and heat transfer coefficients, achieving errors within 0.39–5.57% and correlation coefficients up to 99.9%.
[0008] Moreover, hybrid frameworks have been extended to handle optimization problems. By integrating ANN and genetic algorithms (GA), researchers have tackled friction factor estimation in micro-photocatalytic reactors, illustrating how combining data-driven prediction with heuristic optimization can yield both accurate and computationally efficient design tools.
[0009] Purely data-driven models, though computationally expedient, often suffer from interpretability issues and may behave inconsistently outside the domain of the training data. A promising hybrid direction is embodied by physics-informed neural networks (PINNs), which embed governing PDEs directly into the training loss function ensuring that solutions adhere to underlying physical laws even with sparse or noisy data. PINNs enable mesh-free approximation of solutions, automatic differentiation, and flexible evaluation across variable grid resolutions.
[0010] These methods empower neural networks to approximate fluid behavior while maintaining physical integrity. They have been successfully applied in flow separation studies (e.g., boundary layer detachment and reattachment), and exhibit improved convergence and computational efficiency relative to traditional CFD. Nevertheless, challenges remain in accurately modeling turbulent, unsteady, and shock-laden flows.
[0011] A generalized framework for CFD-ML integration leverages Python libraries like CFFI and dynamic linking to facilitate data transfer and bidirectional coupling in simulations providing a scalable, flexible, and language-agnostic foundation for hybrid modeling workflows on heterogeneous HPC platforms.
[0012] Generative models, including AEs and GANs, similarly facilitate direct prediction of 2D/3D flow fields enabling compact, efficient representations that support rapid inference and enable real-time or near-real-time analysis.
[0013] Hardware-level strategies have further enhanced the efficiency of hybrid CFD/ML systems. Low-precision (e.g., 16-bit floating point) inference on AI accelerators significantly reduces computation time while maintaining acceptable accuracy. Parallelization approaches, such as domain decomposition combined with neighbor-buffer data exchange, enable scalable distributed computations integral for real-time fluid analysis frameworks.
[0014] Thus, in light of the above-stated discussion, there exists a need for a hybrid CFD-ANN framework system for real-time fluid flow analysis.
SUMMARY OF THE DISCLOSURE
[0015] The following is a summary description of illustrative embodiments of the invention. It is provided as a preface to assist those skilled in the art to more rapidly assimilate the detailed design discussion which ensues and is not intended in any way to limit the scope of the claims which are appended hereto in order to particularly point out the invention.
[0016] According to illustrative embodiments, the present disclosure focuses on a hybrid CFD-ANN framework system for real-time fluid flow analysis which overcomes the above-mentioned disadvantages or provide the users with a useful or commercial choice.
[0017] An objective of the present disclosure is to establish a foundation for future extensions of hybrid AI-physics systems in other computationally intensive domains beyond fluid dynamics.
[0018] Another objective of the present disclosure is to develop a hybrid framework that integrates the accuracy of Computational Fluid Dynamics (CFD) with the computational efficiency of Artificial Neural Networks (ANN) for real-time fluid flow analysis.
[0019] Another objective of the present disclosure is to reduce computational time required for fluid flow simulations in complex geometries without compromising the precision of results.
[0020] Another objective of the present disclosure is to enable adaptive, online monitoring and control of dynamic fluid systems by providing near-instantaneous predictions of flow characteristics.
[0021] Another objective of the present disclosure is to design an ANN model trained on CFD-generated datasets that can approximate fluid flow behavior with minimal error.
[0022] Another objective of the present disclosure is to enhance the generalization ability of ANN models by embedding physics-informed constraints derived from CFD principles.
[0023] Another objective of the present disclosure is to provide a scalable solution capable of handling high-resolution meshes and diverse flow conditions in real-time.
[0024] Another objective of the present disclosure is to improve the feasibility of rapid design iterations in engineering applications such as aerospace, automotive, and biomedical fluid dynamics.
[0025] Another objective of the present disclosure is to validate the hybrid CFD-ANN framework against benchmark fluid flow problems and experimentally measured datasets.
[0026] Yet another objective of the present disclosure is to optimize the balance between computational accuracy and speed, ensuring the framework is suitable for both research and industrial applications.
[0027] In light of the above, a hybrid CFD-ANN framework system for real-time fluid flow analysis comprises a computational fluid dynamics (CFD) solver module configured to generate baseline fluid flow solutions for steady, unsteady, laminar, turbulent, single-phase, and multiphase flows. The system also includes an artificial neural network (ANN) module configured to predict fluid flow parameters in real-time by performing computationally efficient approximations. The system also includes a coupling controller module configured to dynamically switch between ANN-based predictions and CFD-based corrections. The system also includes a real-time processing interface configured to integrate outputs from the ANN module and the CFD solver module to provide adaptive fluid flow analysis across a plurality of engineering applications. The system also includes a communication and visualization layer configured to present real-time analysis results to external systems, user interfaces, or control environments.
[0028] In one embodiment, the CFD solver module is further configured to implement numerical methods selected from finite volume, finite element, and lattice Boltzmann methods for generating baseline fluid flow solutions.
[0029] In one embodiment, the CFD solver module supports analysis of multiphase flows comprising combinations of gas-liquid, liquid-liquid, and gas-liquid-solid phases.
[0030] In one embodiment, the ANN module comprises a deep neural network trained on CFD-generated datasets to predict fluid flow parameters including velocity, pressure, temperature, and turbulence characteristics in real-time.
[0031] In one embodiment, the ANN module includes uncertainty quantification capabilities to estimate confidence intervals of predicted fluid flow parameters.
[0032] In one embodiment, the ANN module is periodically retrained using newly generated CFD data to maintain predictive accuracy across varying flow conditions.
[0033] In one embodiment, the coupling controller module implements a dynamic scheduling algorithm to optimize computational resource allocation between the CFD solver and the ANN module.
[0034] In one embodiment, the coupling controller module is configured to initiate CFD-based corrections based on a predefined error threshold between ANN predictions and CFD solutions.
[0035] In one embodiment, the real-time processing interface is further configured to integrate outputs from the ANN module and the CFD solver module using weighted averaging or adaptive fusion techniques to ensure accurate and consistent fluid flow analysis.
[0036] In one embodiment, the communication and visualization layer provide graphical visualization of fluid flow parameters in two-dimensional and three-dimensional formats, along with real-time alerts for deviations or anomalies.
[0037] These and other advantages will be apparent from the present application of the embodiments described herein.
[0038] The preceding is a simplified summary to provide an understanding of some embodiments of the present invention. This summary is neither an extensive nor exhaustive overview of the present invention and its various embodiments. The summary presents selected concepts of the embodiments of the present invention in a simplified form as an introduction to the more detailed description presented below. As will be appreciated, other embodiments of the present invention are possible utilizing, alone or in combination, one or more of the features set forth above or described in detail below.
[0039] These elements, together with the other aspects of the present disclosure and various features are pointed out with particularity in the claims annexed hereto and form a part of the present disclosure. For a better understanding of the present disclosure, its operating advantages, and the specified object attained by its uses, reference should be made to the accompanying drawings and descriptive matter in which there are illustrated exemplary embodiments of the present disclosure.
BRIEF DESCRIPTION OF THE DRAWINGS
[0040] To describe the technical solutions in the embodiments of the present disclosure or in the prior art more clearly, the following briefly describes the accompanying drawings required for describing the embodiments or the prior art. Apparently, the accompanying drawings in the following description merely show some embodiments of the present disclosure, and a person of ordinary skill in the art can derive other implementations from these accompanying drawings without creative efforts. All of the embodiments or the implementations shall fall within the protection scope of the present disclosure.
[0041] The advantages and features of the present disclosure will become better understood with reference to the following detailed description taken in conjunction with the accompanying drawing, in which:
[0042] FIG. 1 illustrates a flowchart outlining sequential step involved in a hybrid CFD-ANN framework system for real-time fluid flow analysis, in accordance with an exemplary embodiment of the present disclosure;
[0043] FIG. 2 illustrates a block diagram of a hybrid CFD-ANN framework system for real-time fluid flow analysis, in accordance with an exemplary embodiment of the present disclosure.
[0044] Like reference, numerals refer to like parts throughout the description of several views of the drawing;
[0045] The hybrid CFD-ANN framework system for real-time fluid flow analysis, which like reference letters indicate corresponding parts in the various figures. It should be noted that the accompanying figure is intended to present illustrations of exemplary embodiments of the present disclosure. This figure is not intended to limit the scope of the present disclosure. It should also be noted that the accompanying figure is not necessarily drawn to scale.
DETAILED DESCRIPTION OF THE DISCLOSURE
[0046] The following is a detailed description of embodiments of the disclosure depicted in the accompanying drawings. The embodiments are in such detail as to communicate the disclosure. However, the amount of detail offered is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure.
[0047] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be apparent to one skilled in the art that embodiments of the present disclosure may be practiced without some of these specific details.
[0048] Various terms as used herein are shown below. To the extent a term is used, it should be given the broadest definition persons in the pertinent art have given that term as reflected in printed publications and issued patents at the time of filing.
[0049] The terms “a” and “an” herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced items.
[0050] The terms “having”, “comprising”, “including”, and variations thereof signify the presence of a component.
[0051] Referring now to FIG. 1 to FIG. 2 to describe various exemplary embodiments of the present disclosure. FIG. 1 illustrates a flowchart outlining sequential step involved in a hybrid CFD-ANN framework system for real-time fluid flow analysis, in accordance with an exemplary embodiment of the present disclosure.
[0052] A hybrid CFD-ANN framework system for real-time fluid flow analysis 100 comprises a computational fluid dynamics (CFD) solver module 102 configured to generate baseline fluid flow solutions for steady, unsteady, laminar, turbulent, single-phase, and multiphase flows. The CFD solver module 102 is further configured to implement numerical methods selected from finite volume, finite element, and lattice Boltzmann methods for generating baseline fluid flow solutions. The CFD solver module 102 supports analysis of multiphase flows comprising combinations of gas-liquid, liquid-liquid, and gas-liquid-solid phases.
[0053] The system also includes an artificial neural network (ANN) module 104 configured to predict fluid flow parameters in real-time by performing computationally efficient approximations. The ANN module 104 comprises a deep neural network trained on CFD-generated datasets to predict fluid flow parameters including velocity, pressure, temperature, and turbulence characteristics in real-time. The ANN module 104 includes uncertainty quantification capabilities to estimate confidence intervals of predicted fluid flow parameters. The ANN module 104 is periodically retrained using newly generated CFD data to maintain predictive accuracy across varying flow conditions.
[0054] The system also includes a coupling controller module 106 configured to dynamically switch between ANN-based predictions and CFD-based corrections. The coupling controller module 106 implements a dynamic scheduling algorithm to optimize computational resource allocation between the CFD solver and the ANN module. The coupling controller module 106 is configured to initiate CFD-based corrections based on a predefined error threshold between ANN predictions and CFD solutions.
[0055] The system also includes a real-time processing interface 108 configured to integrate outputs from the ANN module and the CFD solver module to provide adaptive fluid flow analysis across a plurality of engineering applications. The real-time processing interface 108 is further configured to integrate outputs from the ANN module and the CFD solver module using weighted averaging or adaptive fusion techniques to ensure accurate and consistent fluid flow analysis.
[0056] The system also includes a communication and visualization layer 110 configured to present real-time analysis results to external systems, user interfaces, or control environments. The communication and visualization layer 110 provide graphical visualization of fluid flow parameters in two-dimensional and three-dimensional formats, along with real-time alerts for deviations or anomalies.
[0057] FIG. 1 illustrates a flowchart outlining sequential step involved in a hybrid CFD-ANN framework system for real-time fluid flow analysis.
[0058] At 102, the process begins with the computational fluid dynamics (CFD) solver module, which is responsible for generating baseline fluid flow solutions. This module handles a wide range of flow conditions, including steady and unsteady flows, laminar and turbulent regimes, as well as single-phase and multiphase scenarios. By solving the governing fluid dynamics equations, the CFD solver establishes a foundation of highly accurate reference data that guides subsequent predictive operations within the system.
[0059] At 104, once the baseline solutions are established, the artificial neural network (ANN) module takes over the majority of real-time computations. The ANN module is trained on datasets generated from prior CFD simulations, allowing it to efficiently predict fluid flow parameters under varying conditions without the computational overhead of repeatedly solving complex fluid dynamics equations. By performing these approximations, the ANN module accelerates the analysis process, enabling near-instantaneous response times which are critical for real-time applications such as engineering design iterations, process control, and interactive simulations.
[0060] At 106, the system’s coupling controller module serves as the intelligent intermediary between the ANN and CFD modules. This controller dynamically monitors the accuracy of the ANN predictions and determines when CFD-based corrections are required. Corrections are triggered either periodically or whenever the prediction deviations exceed a pre-defined threshold. This dynamic switching ensures that the system maintains high accuracy without sacrificing the computational efficiency offered by the ANN, creating a robust hybrid approach that leverages the strengths of both methodologies.
[0061] At 108, after predictions and corrections are generated, the real-time processing interface integrates the outputs from both the ANN and CFD modules. This interface consolidates the results into coherent, adaptive fluid flow analyses that can be applied across multiple engineering applications. By intelligently merging fast ANN approximations with precise CFD corrections, the interface ensures that the final output is both timely and reliable, catering to diverse operational requirements such as aerospace design, chemical processing, and environmental simulations.
[0062] At 110, the communication and visualization layer presents the real-time analysis results to end users or connected systems. This layer translates complex computational outputs into interpretable visualizations and data streams, facilitating user interaction, monitoring, and control. It allows engineers or automated systems to make informed decisions based on up-to-date fluid flow information, thereby enhancing responsiveness and situational awareness. Overall, the flowchart demonstrates a seamless interplay between high-fidelity computation and fast predictive modeling, establishing a system capable of delivering real-time, accurate, and adaptive fluid flow analysis in a variety of complex scenarios.
[0063] FIG. 2 illustrates a block diagram of a hybrid CFD-ANN framework system for real-time fluid flow analysis.
[0064] The process begins with the CFD module. This module is responsible for performing high-fidelity simulations of fluid flows, providing accurate baseline solutions that account for a wide range of conditions such as steady, unsteady, laminar, turbulent, single-phase, and multiphase flows. The CFD solver applies the governing equations of fluid dynamics to generate precise flow parameter data, which serves as the foundation for the subsequent predictive operations within the system.
[0065] From the CFD module, data is fed both to the ANN module and to another instance of the CFD module in the second row. The ANN module, positioned at the top right, leverages machine learning techniques to predict fluid flow parameters in real-time, using patterns learned from previously computed CFD datasets. By performing these approximations, the ANN reduces computational load, allowing the system to rapidly estimate flow characteristics while maintaining a close correlation with high-fidelity CFD outputs. The ANN output is then directed to the visualization and interface layer, where the real-time predictions are integrated, processed, and presented in a format suitable for user interaction or external system communication.
[0066] Meanwhile, the second instance of the CFD module in the middle row acts as a corrective mechanism. It receives data from the initial CFD computations as well as feedback from the ANN outputs via the interface layer, allowing the system to reconcile discrepancies and refine predictions as needed. This iterative process ensures that the hybrid framework maintains both speed and accuracy, with the ANN handling most of the routine computations and the CFD module providing periodic high-accuracy corrections.
[0067] Finally, the data manager, located at the bottom of the flowchart, consolidates outputs from the CFD computations and the predictive ANN module. It ensures that all fluid flow data is systematically stored, organized, and made available for further analysis, validation, or integration with other engineering systems. The combination of CFD-based accuracy, ANN-driven computational efficiency, and robust data management forms a cohesive system capable of delivering real-time, adaptive, and reliable fluid flow analysis for a wide variety of engineering applications.
[0068] While the invention has been described in connection with what is presently considered to be the most practical and various embodiments, it will be understood that the invention is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims.
[0069] A person of ordinary skill in the art may be aware that, in combination with the examples described in the embodiments disclosed in this specification, units and algorithm steps may be implemented by electronic hardware, computer software, or a combination thereof.
[0070] The foregoing descriptions of specific embodiments of the present disclosure have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed, and many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described to best explain the principles of the present disclosure and its practical application, and to thereby enable others skilled in the art to best utilize the present disclosure and various embodiments with various modifications as are suited to the particular use contemplated. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but such omissions and substitutions are intended to cover the application or implementation without departing from the scope of the present disclosure.
[0071] Disjunctive language such as the phrase “at least one of X, Y, Z,” unless specifically stated otherwise, is otherwise understood with the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and/or Z). Thus, such disjunctive language is not generally intended to, and should not, imply that certain embodiments require at least one of X, at least one of Y, or at least one of Z to each be present.
[0072] In a case that no conflict occurs, the embodiments in the present disclosure and the features in the embodiments may be mutually combined. The foregoing descriptions are merely specific implementations of the present disclosure, but are not intended to limit the protection scope of the present disclosure. Any variation or replacement readily figured out by a person skilled in the art within the technical scope disclosed in the present disclosure shall fall within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.
, Claims:I/We Claim:
1. A hybrid CFD-ANN framework system for real-time fluid flow analysis (100) comprising:
a computational fluid dynamics (CFD) solver module (102) configured to generate baseline fluid flow solutions for steady, unsteady, laminar, turbulent, single-phase, and multiphase flows;
an artificial neural network (ANN) module (104) configured to predict fluid flow parameters in real-time by performing computationally efficient approximations;
a coupling controller module (106) configured to dynamically switch between ANN-based predictions and CFD-based corrections;
a real-time processing interface (108) configured to integrate outputs from the ANN module and the CFD solver module to provide adaptive fluid flow analysis across a plurality of engineering applications;
a communication and visualization layer (110) configured to present real-time analysis results to external systems, user interfaces, or control environments.
2. The system (100) as claimed in claim 1, wherein the CFD solver module (102) is further configured to implement numerical methods selected from finite volume, finite element, and lattice Boltzmann methods for generating baseline fluid flow solutions.
3. The system (100) as claimed in claim 1, wherein the CFD solver module (102) supports analysis of multiphase flows comprising combinations of gas-liquid, liquid-liquid, and gas-liquid-solid phases.
4. The system (100) as claimed in claim 1, wherein the ANN module (104) comprises a deep neural network trained on CFD-generated datasets to predict fluid flow parameters including velocity, pressure, temperature, and turbulence characteristics in real-time.
5. The system (100) as claimed in claim 1, wherein the ANN module (104) includes uncertainty quantification capabilities to estimate confidence intervals of predicted fluid flow parameters.
6. The system (100) as claimed in claim 1, wherein the ANN module (104) is periodically retrained using newly generated CFD data to maintain predictive accuracy across varying flow conditions.
7. The system (100) as claimed in claim 1, wherein the coupling controller module (106) implements a dynamic scheduling algorithm to optimize computational resource allocation between the CFD solver and the ANN module.
8. The system (100) as claimed in claim 1, wherein the coupling controller module (106) is configured to initiate CFD-based corrections based on a predefined error threshold between ANN predictions and CFD solutions.
9. The system (100) as claimed in claim 1, wherein the real-time processing interface (108) is further configured to integrate outputs from the ANN module and the CFD solver module using weighted averaging or adaptive fusion techniques to ensure accurate and consistent fluid flow analysis.
10. The system (100) as claimed in claim 1, wherein the communication and visualization layer (110) provide graphical visualization of fluid flow parameters in two-dimensional and three-dimensional formats, along with real-time alerts for deviations or anomalies.