Learning Non-Ideal Single Vortex Flows Using the Differentiable Vortex Particle Method
Ziqi Ji, Gang Du, Penghao Duan · 2025
Read Learning Non-Ideal Single Vortex Flows Using the Differentiable Vortex Particle Method by Ziqi Ji, Gang Du, Penghao Duan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-17012.txt.
Introduction
Ziqi Ji, Gang Du, and Penghao Duan extend a learning method for swirling flows beyond ideal cases. Their 2025 paper deals with vortex flows, seen in typhoons, currents, and aircraft wakes. A particle method tracks vortices in such flows. A newer hybrid adds deep learning so flows can be learned and predicted. The hybrid, though, assumed ideal flow: no viscosity and no outside forces.
The team generalizes the method to viscous flow and to flows pushed by non conservative forces. A classic test case anchors the work. The Lamb-Oseen vortex, a fundamental viscous swirl, has an exact solution from the governing equations. That gives firm ground truth for training and checking. Across flow strengths, the method modeled the vortex more accurately than standard convolutional networks. The authors aim at real engineering use.
Catalog record
| CREATOR | Ziqi Ji, Gang Du, Penghao Duan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-17012
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Learning Non-Ideal Single Vortex Flows Using the Differentiable Vortex Particle Method Ziqi Ji, Gang Du, Penghao Duan Abstract This study extends the differentiable vortex particle method (DVPM) beyond idealized flow scenarios to encompass more realistic, non-ideal conditions, including viscous flow and flow subjected to non-conservative body forces. We establish the Lamb-Oseen vortex as a benchmark case, representing a fundamental viscous single vortex flow in fluid mechanics. This selection offers significant analytical advantages, as the Lamb-Oseen vortex possesses an exact analytical solution derived from the Navier-Stokes (NS) equations, thereby providing definitive ground truth data for training and validation purposes. Through rigorous evaluation across a spectrum of Reynolds numbers, we demonstrate that DVPM achieves superior accuracy in modeling the Lamb-Oseen vortex compared to conventional convolutional neural networks (CNNs) and physics-informed neural networks (PINNs). Our results substantiate DVPM’s robust capabilities in modeling non-ideal single vortex flows, establishing its distinct advantages over contemporary deep learning methodologies in fluid dynamics applications. The dataset and source code are publicly available on GitHub at the following link: https://github.com/jh36714753/Learning_Non-Ideal_Single_Vortex_Flows . 1 Introduction Vortex flows represent a fundamental class of fluid phenomena prevalent throughout nature. They manifest in diverse forms, such as turbulent structures, swirling water patterns, turbomachinery flows, and atmospheric systems, including hurricanes. Despite their ubiquity and significant importance, the complete understanding of vortex flow dynamics remains an open challenge in contemporary scientific research. The concept of hidden fluid mechanics was introduced by Raissi et al. [ 8 ] in 2020, referring to fluid mechanical phenomena not explicitly represented by governing equations but instead embedded within flow visualizations. This paradigm aims to decode and learn fluid dynamics directly from visualization data. Currently, there are three principal machine learning approaches to hidden fluid mechanics: the PINNs-based method [ 8 ] , the CNN-based method [ 15 ] , and the DVPM-based approach [ 2 ] . In their pioneering PINNs-based method, Raissi et al. [ 8 ] enhanced the conventional PINNs framework by incorporating an additional scalar variable representing marked material concentration to visualize flow patterns. This methodology demonstrated remarkable success in addressing various physical and biomedical applications. Subsequently, Zhang et al. [ 15 ] developed the CNN-based approach, which integrates physical constraints into the loss function of the CNN framework. Their method effectively utilized particle image velocimetry (PIV) data for flow prediction. More recently, Deng et al. [ 2 ] proposed the DVPM-based method, introducing a deep learning-based vortex particle method specifically optimized for vortex flows. This specialized approach has exhibited superior performance to PINNs and CNN methods in analyzing vortical flow structures. The DVPM, introduced by Deng et al. [ 2 ] , which integrates deep learning techniques with traditional fluid dynamics approaches. This innovative method harnesses the differentiable properties of neural networks to enhance vortex flow prediction and analysis. However, the method’s initial implementation exhibits several limitations that constrain its practical applicability. Specifically, it is restricted to highly simplified vortex flow scenarios, including two-dimensional flows, inviscid conditions, and cases without non-conservative body forces. These constraints significantly limit its extension to more complex, real-world vortex flows. Additionally, a critical consideration arises regarding the method’s training dataset, which partially relies on traditional vortex particle method simulations. From the perspective of fluid mechanics researchers, these simulations may not provide sufficient accuracy to serve as reliable ground truth data. This study extends the differentiable vortex particle method to encompass more general, non-ideal conditions, including viscous flow and flow influenced by non-conservative body forces. As a benchmark case, we introduce the viscous single vortex flow, specifically the Lamb-Oseen vortex. The Lamb-Oseen vortex, with its analytical solution derived directly from the NS equations, provides ground truth data for model training and validation. Through rigorous evaluation across various Reynolds numbers, we demonstrate that the differentiable vortex particle method achieves superior accuracy compared to the CNN-based and PINN-based approaches. Additionally, we validate our model’s capability to accurately predict single vortex flow dynamics under the combined influence of viscosity and non-conservative body forces. These results substantiate the robust performance of the differentiable vortex particle method in vortex flow modeling and establish its significant advantages over contemporary deep learning techniques. 2 Related work The vortex method is a prominent computational technique that has garnered significant attention across multiple domains. Its applications extend beyond computational fluid dynamics (CFD) [ 4 , 5 , 6 , 13 ] to computer graphics (CG) [ 9 , 7 , 12 , 1 , 3 , 14 , 10 , 16 , 11 ] . This methodology demonstrates the inherent advantages of meshless approaches, employs a Lagrangian viewpoint, and exhibits superior computational efficiency. 3 Physical model For incompressible fluid flows, the vorticity transport equation, derived rigorously from the NS equations, establishes a comprehensive mathematical framework for characterizing the spatiotemporal evolution of vorticity in swirling flow fields: \frac{D\boldsymbol{\omega}}{Dt}=\frac{\partial\boldsymbol{\omega}}{\partial t}+\boldsymbol{u}\cdot\nabla\boldsymbol{\omega}=\boldsymbol{\omega}\cdot\nabla\boldsymbol{u}+\nu\nabla^{2}\boldsymbol{\omega}+\nabla\times\boldsymbol{b}, (1) where \boldsymbol{\omega} represents the vorticity vector, t signifies time, \boldsymbol{u} denotes the velocity field, \boldsymbol{b} corresponds to the non-conservative body force, and \nu refers to the kinematic viscosity. For an inviscid, two-dimensional flow in the absence of non-conservative body forces, the equation reduces to D\boldsymbol{\omega}/Dt=\mathbf{0} , which explicitly demonstrates the Lagrangian conservation of vorticity (i.e., the vorticity of a fluid particle remains constant as it is advected along its trajectory). However, practical flows rarely satisfy these idealized conditions, leading to more complex flow phenomena characterized by intricate vortical dynamics. Through the application of the Biot-Savart Law, we derive the vortex-induced velocity formulation: \boldsymbol{u}(\boldsymbol{x})=\int K\left(\boldsymbol{x}-\boldsymbol{x}^{\prime}\right)\omega\left(\boldsymbol{x}^{\prime}\right)d\boldsymbol{x}^{\prime}, (2) where K\left(\boldsymbol{x}-\boldsymbol{x}^{\prime}\right) denotes the kernel function, which is dependent on the displacement vector between the vorticity location and the point of interest in the flow field. In essence, Equation 2 indicates that the velocity at any point within the flow field can be determined by integrating the contributions of vorticity distributed throughout the domain, mediated by the kernel function. Figure 1 presents the physical model employed in our computational framework. Within a two-dimensional flow field comprising n discrete vortices, each vortex is characterized by three fundamental parameters: position \boldsymbol{x}_{i}\prime , vorticity \omega_{i} , and radius \delta_{i} (assuming circular vortex geometry) for the i-th vortex. These vortices utilize a common kernel function K\left(\frac{\boldsymbol{x}-\boldsymbol{x}^{\prime}}{\delta}\right) . Subsequently, the velocity at any arbitrary position \boldsymbol{x} within the flow field can be determined through: \boldsymbol{u}(\boldsymbol{x})=\sum_{i=1}^{n}K\left(\frac{\left|\boldsymbol{x}-\boldsymbol{x}_{i}^{\prime}\right|}{\delta_{i}}\right)\omega\left(\boldsymbol{x}_{i}^{\prime}\right) (3) By Equation 3 , estimating velocity throughout the flow fields necessitates knowledge of three key attributes of all vortices, along with their kernel function. To address this computational challenge, we implement a framework comprising four neural networks, each dedicated to predicting one of these essential physical quantities. In the two-dimensional flows, the temporal evolution of positional coordinates for n vortices can be expressed as \left[\left(\boldsymbol{x}_{1}\right)_{t},\ldots,\left(\boldsymbol{x}_{n}\right)_{t}\right] , while their vorticity distributions are represented by \left[\left(\omega_{1}\right)_{t},\ldots,\left(\omega_{n}\right)_{t}\right] , and their radial parameters are denoted by \left[\left(\delta_{1}\right)_{t},\ldots,\left(\delta_{n}\right)_{t}\right] . Consequently, we employ four distinct neural networks to characterize these fundamental physical quantities: \left\{\begin{array}[]{l}N_{\boldsymbol{x}}(t)=\left[\left(\boldsymbol{x}_{1}\right)_{t},\cdots,\left(\boldsymbol{x}_{n}\right)_{t}\right]\\ N_{\omega}(t)=\left[\left(\omega_{1}\right)_{t},\cdots,\left(\omega_{n}\right)_{t}\right]\\ N_{\delta^{2}}(t)=\left[\left(\delta_{1}^{2}\right)_{t},\cdots,\left(\delta_{n}^{2}\right)_{t}\right]\\ N_{K}\left(\frac{\left|\boldsymbol{x}_{i}-\boldsymbol{x}_{i}^{\prime}\right|}{\delta_{i}}\right)=K\left(\frac{\left|\boldsymbol{x}_{i}-\boldsymbol{x}_{i}^{\prime}\right|}{\delta_{i}}\right)\end{array}\right. (4) Figure 1: The physical model in our framework. K\left(\left|\boldsymbol{\boldsymbol{x}}-\boldsymbol{x}\prime\right|/\delta\right) represents the kernel function, whose value is determined by multiple parameters: the dependent variable, the Euclidean distance between the vortex core and the point of interest in the flow field \left|\boldsymbol{x}-\boldsymbol{x}_{p}\right| , and the vortex radius \delta . Each vortex is characterized by a distinct vorticity \omega , which reduces to a scalar quantity in