arXiv subjects

Marie-Christine Volk

Publications and source records attributed to Marie-Christine Volk.

4 recordsLinked to original sources

arXiv2026

An improved periodic activation for PINNs reconstructing convective flows

Architectures with periodic activation functions have already been shown to be beneficial in comparison to monotonic counterparts for a wide range of applications of physics-informed neural networks. Here, we investigate a network architecture which uses the complex exponential function, generating pairs of sine and cosine outputs as activation functions. Testing it against comparable, sine-activated multi-layer perceptrons for the task of temperature reconstruction from sparse velocity data for cubic Rayleigh-Bénard convection reveals significant improvements in the reconstruction quality without a substantial increase in computational cost per training step. Vice versa, the improved architecture enables reaching similar reconstruction qualities for a fraction of the expense. Analyzing the mathematical structure of these networks points to the improvements being rooted in the property of passing both a sine and cosine function forward. This way, the subsequent layer is able to adapt the phase of the provided latent periodic functions, and doing it individually for each of its neurons.

physics.flu-dyn↗

arXiv2025

Curvature-based energy spectra revealing flow regime changes in Rayleigh-Bénard convection

We use the local curvature derived from velocity vector fields or particle tracks as a surrogate for structure size to compute curvature-based energy spectra. An application to homogeneous isotropic turbulence shows that these spectra replicate certain features of classical energy spectra such as the slope of the inertial range extending towards the equivalent curvature of the Taylor microscale. As this curvature-based analysis framework is sampling based, it also allows further statistical analyses of the time evolution of the kinetic energies and curvatures considered. The main findings of these analyses are that the slope for the inertial range also appears as a salient point in the probability density distribution of the angle of the vector comprising the two time evolution components. This density distribution further exhibits changing features of its shape depending on the Rayleigh number. This Rayleigh number evolution allows to observe a change in the flow regime between the Rayleigh numbers $10^6$ and $10^7$. Insight into this regime change is gathered by conditionally sampling the salient time evolution behaviours and projecting them back into physical space. Concretely, the regime change is manifested by a change in the spatial distribution for the different time evolution behaviours. Finally, we show that this analysis can be applied to measured Lagrangian particle tracks.

physics.flu-dyn↗

arXiv2025

A PINN Methodology for Temperature Field Reconstruction in the PIV Measurement Plane: Case of Rayleigh-Bénard Convection

We present a method to infer temperature fields from stereo particle-image velocimetry (PIV) data in turbulent Rayleigh-Bénard convection (RBC) using Physics-informed neural networks (PINNs). The physical setup is a cubic RBC cell with Rayleigh number $\text{Ra}=10^7$ and Prandtl number $\text{Pr}=0.7$. With data only available in a vertical plane $A:x=x_0$, the residuals of the governing partial differential equations are minimised in an enclosing 3D domain around $A$ with thickness $δ_x$. Dynamic collocation point sampling strategies are used to overcome the lack of 3D labelled information and to optimize the overall convergence of the PINN. In particular, in the out-of-plane direction $x$, the collocation points are distributed according to a normal distribution, in order to emphasize the region where data is provided. Along the vertical direction, we leverage meshing information and sample points from a distribution designed based on the grid of a direct numerical simulation (DNS). This approach points greater attention to critical regions, particularly the areas with high temperature gradients within the thermal boundary layers. Using planar three-component velocity data from a DNS, we successfully validate the reconstruction of the temperature fields in the PIV plane. We evaluate the robustness of our method with respect to characteristics of the labelled data used for training: the data time span, the sampling frequency, some noisy data and boundary data omission, aiming to better accommodate the challenges associated with experimental data. Developing PINNs on controlled simulation data is a crucial step toward their effective deployment on experimental data. The key is to systematically introduce noise, gaps, and uncertainties in simulated data to mimic real-world conditions and ensure robust generalization.

physics.flu-dyn↗

arXiv2024

Periodically activated physics-informed neural networks for assimilation tasks for three-dimensional Rayleigh-Bénard convection

We apply physics-informed neural networks to three-dimensional Rayleigh-Bénard convection in a cubic cell with a Rayleigh number of Ra = 10^6 and a Prandtl number of Pr = 0.7 to assimilate the velocity vector field from given temperature fields and vice versa. With the respective ground truth data provided by a direct numerical simulation, we are able to evaluate the performance of the different activation functions applied (sine, hyperbolic tangent and exponential linear unit) and different numbers of neurons (32, 64, 128, 256) for each of the five hidden layers of the multi-layer perceptron. The main result is that the use of a periodic activation function (sine) typically benefits the assimilation performance in terms of the analyzed metrics, correlation with the ground truth and mean average error. The higher quality of results from sine-activated physics-informed neural networks is also manifested in the probability density function and power spectra of the inferred velocity or temperature fields. Regarding the two assimilation directions, the assimilation of temperature fields based on velocities appears to be more challenging in the sense that it exhibits a sharper limit on the number of neurons below which viable assimilation results can not be achieved.

physics.flu-dyn↗