Computers & Fluids最新文献

{"title":"An enhanced second-order discretization scheme for free surface and vicinity particles in MPS method aided by surface mesh","authors":"Gen Li ,&nbsp;Yunlong Liao ,&nbsp;Peitao Yao","doi":"10.1016/j.compfluid.2026.106996","DOIUrl":"10.1016/j.compfluid.2026.106996","url":null,"abstract":"<div><div>Moving Particle Semi-implicit (MPS) method is an emerging numerical method for free-surface flow involving complex deformation, fragmentation and coalescence of various fluid interfaces. However, higher-order discretization schemes for the MPS method remain imperfect. The non-uniform particle distribution near the free surface is highly prone to causing numerical divergence. The conventional virtual-particle-based second-order discretization scheme degrades the discretization scheme for the free surface and particles in its vicinity to a lower-order format. This treatment thus leads to error accumulation and propagation. To solve these problems, an improved second-order discretization scheme was developed with the aid of a surface mesh. A surface mesh constructed at the free surface provided the missing position information for neighboring particles required by the surface particles’ second-order discretization and compensated for particle number density deficiencies. A sensitivity analysis was conducted on surface mesh resolution and the particle-to-mesh size ratio was determined to balance computational efficiency and accuracy. Compared to the prior virtual-particle-based second-order method, the proposed approach enabled accurate discretization for free surface particles, preventing error accumulation caused by non-uniform particle distributions. Validations were conducted by simulating four benchmark cases of still water pool pressure, dam break flow, elliptical droplet evolution, and square droplet rotation. The results demonstrated that the proposed surface-mesh-based method exhibited superior performance in pressure calculation accuracy, free surface particle distribution uniformity, and surface consistency.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"308 ","pages":"Article 106996"},"PeriodicalIF":3.0,"publicationDate":"2026-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146076191","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Prediction of surface pressure distributions of non-parametric airfoils using geometric deep learning methods","authors":"Derrick Hines ,&nbsp;Philipp Bekemeyer","doi":"10.1016/j.compfluid.2026.106979","DOIUrl":"10.1016/j.compfluid.2026.106979","url":null,"abstract":"<div><div>Computational Fluid Dynamics (CFD) simulations are one of the cornerstones in providing aerodynamic data required for aircraft design and optimization. However, using these simulations extensively is limited by their high computational demands. Therefore, it is essential to create efficient data-driven surrogate models for CFD solvers. In many practical scenarios an explicit unifying parameterization of aircraft configurations is not available. This highlights the need for models that operate directly on raw geometric representations. Geometric deep learning has emerged as a class of deep learning techniques capable of operating on such data, enabling predictive modeling without the need of an explicit parameterization. In this paper, we extend and investigate two geometric deep learning approaches for the prediction of surface pressure distributions of non-parametric airfoils. These methods are Bi-Stride Multi-Scale Graph Neural Network and Implicit Neural Representation of the signed distance function coupled with a Multi-Layer Perceptron. To enhance both of these methods, we propose the use of area-weighted loss functions to better account for variations in node density in the meshes. Moreover, in the formulation of the graph neural network we introduce edge completion at the coarsest level to account for interactions between different connected components, such as flap, main element and slat in a 3-element high-lift airfoil. These methods are compared to the well-established method Proper Orthogonal Decomposition coupled with Interpolation, which is allowed to use an explicit parameterization and serves as a baseline. Two high-fidelity datasets with CFD simulations solving the Reynold-Averaged Navier Stokes equations are created. The first one features a varied set of single-element airfoils with simulations in the subsonic and transonic regime, while the second one features high-lift multi-element airfoils with a variable flap position with simulations in the subsonic regime. The results show that both geometric deep learning approaches outperform the established baseline across various data regimes. These approaches can capture shocks and flow separation with more accuracy. The use of an area-weighted loss function enhances area-weighted performance and leads to faster performance gains in the early training epochs. These findings support the potential of geometric deep learning methods as data-driven surrogates of CFD solvers for varying geometries.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"308 ","pages":"Article 106979"},"PeriodicalIF":3.0,"publicationDate":"2026-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146076192","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Burst buffer accelerated direct numerical simulation of turbulence generated by 3D sparse multiscale grids","authors":"Syed M. Usama ,&nbsp;Nadeem A. Malik ,&nbsp;Umair Umer ,&nbsp;Amjad Shaikh ,&nbsp;Zhigang Sun","doi":"10.1016/j.compfluid.2026.106985","DOIUrl":"10.1016/j.compfluid.2026.106985","url":null,"abstract":"<div><div>This study investigates the free-stream turbulence characteristics generated by a novel three-dimensional sparse multiscale grid (3DSG) using a burst buffer accelerated direct numerical simulation algorithm (BBDSA). The BBDSA achieves a fivefold reduction in computation time and an eightfold improvement in parallel efficiency by employing burst buffers to absorb large data volumes, thereby mitigating I/O bottlenecks in parallel file systems and enabling faster computation of turbulent flows. The algorithm was applied to examine turbulence modulation induced by three types of turbulence generating grids: two-dimensional classical, two-dimensional fractal, and 3DSG configurations. Simulations were conducted for a uniform inflow at a Reynolds number of 4000 within a conduit representative of a wind tunnel. Comparative analyses revealed that the 3DSG with a 24% blockage ratio produced turbulence intensities and Reynolds stresses comparable to those generated by classical and fractal grids with substantially higher blockage ratios. These findings advance the understanding of turbulent flow, highlighting the potential of sparse multiscale grids for efficient turbulence production and their applicability in the design of flow-sensitive engineering systems.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"308 ","pages":"Article 106985"},"PeriodicalIF":3.0,"publicationDate":"2026-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146076189","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neural physics: Using AI libraries to develop physics-based solvers for incompressible computational fluid dynamics","authors":"Boyang Chen ,&nbsp;Claire E. Heaney ,&nbsp;Christopher C. Pain","doi":"10.1016/j.compfluid.2026.106981","DOIUrl":"10.1016/j.compfluid.2026.106981","url":null,"abstract":"<div><div>Numerical discretisations of partial differential equations (PDEs) can be written as discrete convolutions, which, themselves, are a key tool in AI libraries and used in convolutional neural networks (CNNs). We therefore propose to implement numerical discretisations as convolutional layers of a neural network, where the weights or filters are determined analytically rather than by training. Furthermore, we demonstrate that these systems can be solved entirely by functions in AI libraries, either by using Jacobi iteration or multigrid methods, the latter realised through a U-Net architecture. Some advantages of the Neural Physics approach are that (1) the methods are platform agnostic; (2) the resulting solvers are fully differentiable, ideal for optimisation tasks; and (3) writing CFD solvers as (untrained) neural networks means that they can be seamlessly integrated with trained neural networks to form hybrid models. We demonstrate the proposed approach on a number of test cases of increasing complexity from advection-diffusion problems, the non-linear Burgers equation to the Navier-Stokes equations. We validate the approach by comparing our results with solutions obtained from traditionally written code and common benchmarks from the literature. We show that the proposed methodology can solve all these problems using repurposed AI libraries in an efficient way, without training, and presents a new avenue to explore in the development of methods to solve PDEs with implicit methods.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"308 ","pages":"Article 106981"},"PeriodicalIF":3.0,"publicationDate":"2026-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146076122","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A coarse-mesh semi-analytical framework for incompressible flows: Extending the Nodal Integral-Immersed Boundary Method","authors":"Amritpal Singh ,&nbsp;Neeraj Kumar ,&nbsp;Abdellah Hadjadj ,&nbsp;Mostafa Safdari Shadloo","doi":"10.1016/j.compfluid.2026.106967","DOIUrl":"10.1016/j.compfluid.2026.106967","url":null,"abstract":"<div><div>This work extends the Nodal Integral-Immersed Boundary Method (NIM-IBM) to the solution of steady incompressible Navier-Stokes equations in complex geometries. The NIM provides a coarse-mesh, semi-analytical discretization that maintains second-order spatial accuracy, while the sharp-interface IBM enforces boundary conditions on non-body-fitted Cartesian grids. To address the challenges of pressure-velocity coupling and mass conservation near immersed boundaries, the formulation integrates a hybrid MAC-SOLA (Marker and Cell-Solution Algorithm) pressure-correction scheme, preserving the analytical structure of NIM and avoiding complex matrix couplings at cut cells. The proposed framework is validated against multiple benchmark problems involving internal and external flows. Results show that the method accurately captures key flow features and benchmark quantities, even on coarse meshes, with good agreement to experimental and high-resolution numerical data. The approach offers a computationally efficient and geometrically flexible alternative for incompressible flow simulations, with potential for extension to unsteady and high-Reynolds-number regimes.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"307 ","pages":"Article 106967"},"PeriodicalIF":3.0,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Diffuse-interface modeling of two-phase flows with a Boussinesq-Scriven interface","authors":"Jang Min Park","doi":"10.1016/j.compfluid.2026.106970","DOIUrl":"10.1016/j.compfluid.2026.106970","url":null,"abstract":"<div><div>In this study, the diffuse-interface method is employed to model a Boussinesq-Scriven interface that incorporates surface viscosity alongside surface tension. This method differs from the sharp-interface approach in its continuous treatment of the additional surface stress in the momentum conservation equation. The finite element formulation and numerical results are presented. Convergence tests are carried out by using the method of manufactured solution, and optimal convergence rates are observed in both time and space. For a two-dimensional droplet deformation problem, the results show that the diffuse-interface method converges to the sharp-interface method as the diffuse-interface thickness decreases. The present formulation is also applied to a two-dimensional droplet coalescence problem to investigate the effect of surface viscosity.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"307 ","pages":"Article 106970"},"PeriodicalIF":3.0,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145974153","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Predicting flow-induced vibration in isolated and tandem cylinders using hypergraph neural networks","authors":"Shayan Heydari,&nbsp;Rui Gao,&nbsp;Rajeev K. Jaiman","doi":"10.1016/j.compfluid.2025.106930","DOIUrl":"10.1016/j.compfluid.2025.106930","url":null,"abstract":"<div><div>We present a finite element-inspired hypergraph neural network framework for predicting flow-induced vibrations in freely oscillating cylinders. The surrogate architecture transforms unstructured computational meshes into node-element hypergraphs that encode higher-order spatial relationships through element-based connectivity, preserving the geometric and topological structure of the underlying finite-element discretization. The temporal evolution of the fluid-structure interaction is modeled via a modular partitioned architecture: a complex-valued, proper orthogonal decomposition-based sub-network predicts mesh deformation using a low-rank representation of Arbitrary Lagrangian-Eulerian (ALE) grid displacements, while a hypergraph-based message-passing network predicts the unsteady flow field using geometry-aware node, element, and hybrid edge features. High-fidelity ALE-based simulations provide training and evaluation data across a range of Reynolds numbers and reduced velocities for isolated and tandem cylinder configurations. The framework demonstrates stable rollouts and accurately captures the nonlinear variation of oscillation amplitudes with respect to reduced velocity, a key challenge in surrogate modeling of flow-induced vibrations. In the tandem configuration, the model successfully resolves complex wake-body interactions and multi-scale coupling effects, enabling prediction of pressure and velocity fields under strong wake interference conditions. Our results show high fidelity in reproducing force statistics, dominant frequencies, and flow-field dynamics, supporting the framework’s potential as a robust surrogate model for digital twin applications.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"307 ","pages":"Article 106930"},"PeriodicalIF":3.0,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883380","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Energy-based feature extraction with adaptive local domain decomposition for prediction of transient and turbulence flow with operator regression models","authors":"Wenzhuo Xu,&nbsp;Madhav Karthikeyakannan,&nbsp;Christopher McComb,&nbsp;Noelia Grande Gutiérrez","doi":"10.1016/j.compfluid.2025.106958","DOIUrl":"10.1016/j.compfluid.2025.106958","url":null,"abstract":"<div><div>Machine learning (ML) based surrogate models offer the potential to accelerate real-world engineering simulations involving millions of elements by bypassing the need for full-scale numerical simulations. However, current model capacities and available GPU memory often impose severe constraints, limiting our ability to accurately represent the highly variant physical dynamics encountered in complex systems. In traditional numerical methods, these computational limitations are mitigated using domain decomposition. The computational domain is split up to enable parallelization of the computation and reduce memory load. Similarly, ML models can benefit from decomposing the domain into subdomains. However, domain decomposition alone is insufficient to guarantee model performance and accuracy when physical dynamics vary spatially. We introduce the Adaptive Local Domain Decomposition (ALDD) method, which features two key innovations. First, it utilizes domain decomposition to improve the training efficiency of the ML model, with time reduction scaling almost linearly with the number of parallel GPUs. Second, ALDD adaptively partitions the domain and schedules appropriate models by segmenting the physics domain into subdomains based on physical dynamics features. Different ML models explicitly trained to solve different physical dynamics are then strategically assigned to these subdomains, encoding boundary information to ensure a smooth transition at the subdomain interface. This is accomplished by analyzing the energy spectrum of each subdomain and applying k-means clustering on the Wasserstein distances to identify physically coherent regions. We demonstrate superior performance and accuracy compared to baseline ML surrogate models for transitional boundary layer flow and recurrent temporal predictions with over 6 million elements.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"307 ","pages":"Article 106958"},"PeriodicalIF":3.0,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145974156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A sixth-order WCNS based on nonpolynomial interpolation with enhanced accuracy and resolution","authors":"Shaoqiang Han ,&nbsp;Xiaogang Deng ,&nbsp;Wenping Song ,&nbsp;Zhonghua Han","doi":"10.1016/j.compfluid.2026.106974","DOIUrl":"10.1016/j.compfluid.2026.106974","url":null,"abstract":"<div><div>The classic fifth-order weighted compact nonlinear scheme (WCNS) suffers from excessive numerical dissipation and an accuracy mismatch between its nonlinear interpolation and flux differences. Although the sixth-order central/upwind WCNS (WCNS-CU6) resolves the accuracy mismatch, it compromises stability. In this paper, an alternative sixth-order WCNS based on nonpolynomial interpolation (WCNS-NP6) is proposed to enhance accuracy and resolution while maintaining stability. The basic framework of WCNS-NP6 relies on the nonlinear weighting of three-point substencils, similar to the classic fifth-order WCNS. However, in WCNS-NP6, a radial basis function (RBF) is used to interpolate variables from point-based stencils to midpoints, and information from a global six-point stencil is integrated through the shape parameter of the RBF to achieve sixth-order accuracy. A novel measurement function is constructed to assess the smoothness of the six-point stencil. Near discontinuities, the measurement function adaptively removes the shape parameter, reverting WCNS-NP6 to the classic fifth-order WCNS and thereby ensuring stability. In smooth regions, the measurement function confines the active range of the nonlinear weights, thereby mitigating the impact of nonlinear mechanisms on spectral properties. Furthermore, a stencil rotation method is presented to ensure that WCNS-NP6 maintains its nominal sixth-order accuracy for solutions containing arbitrary numbers and orders of critical points. The numerical tests demonstrate that WCNS-NP6 outperforms classic fifth-order and sixth-order WCNSs in terms of numerical dissipation, resolution, and accuracy, particularly at high-order critical points. Notably, the WCNS-NP6 scheme demonstrates better stability than the classical sixth-order WCNS-CU6 scheme, while the computational cost increases by only 19% in 2D benchmark inviscid cases and remains below 10% in a 3D viscous case in engineering.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"307 ","pages":"Article 106974"},"PeriodicalIF":3.0,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145974157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gauss-Newton Natural Gradient Descent for Physics-informed Computational Fluid Dynamics","authors":"Anas Jnini ,&nbsp;Flavio Vella ,&nbsp;Marius Zeinhofer","doi":"10.1016/j.compfluid.2025.106955","DOIUrl":"10.1016/j.compfluid.2025.106955","url":null,"abstract":"<div><div>We propose Gauss-Newton’s method in function space for the solution of the Navier-Stokes equations in the physics-informed neural network (PINN) framework. Upon discretization, this yields a natural gradient method that provably mimics the function space dynamics. Our computational results demonstrate close to single-precision accuracy measured in relative <em>L</em>² norm on a number of benchmark problems. To the best of our knowledge, this constitutes the first contribution in the PINN literature that solves the Navier-Stokes equations to this degree of accuracy. Finally, we show that given a suitable integral discretization, the proposed optimization algorithm agrees with Gauss-Newton’s method in parameter space. This allows a matrix-free formulation enabling efficient scalability to large network sizes.</div></div>","PeriodicalId":287,"journal":{"name":"Computers & Fluids","volume":"307 ","pages":"Article 106955"},"PeriodicalIF":3.0,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940422","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}