arXiv subjects

Alexander Stroh

Publications and source records attributed to Alexander Stroh.

10 recordsLinked to original sources

arXiv2025

Data-driven correlations for thermohydraulic roughness properties

The influence of rough surfaces on fluid flow is characterized by the downward shift in the logarithmic layer of velocity and temperature profiles, namely the velocity roughness function $\Delta U^+$ and the corresponding temperature roughness function $\Delta \Theta^+$. Their computation relies on computational simulations, and hence a simple prediction without such simulation is envisioned. We present a framework, where a data-driven model is developed using the dataset of Yang et al. 2023 \cite{yang_2023} with $93$ high fidelity direct numerical simulations of a fully-developed turbulent channel flow at $Re_\tau \approx 800$ and $Pr = 0.71$. The model provides robust predictive capabilities (mean squared error $\text{MSE}_{k} = 0.09$ and $\text{MSE}_\theta = 0.096$), but lacks interpretability. Simplistic statistical roughness parameters provide a more understandable route, so the framework is extended with a symbolic regression approach to distill an empirical correlation from the data-driven model. The derived expression leads to a predictive correlation for the equivalent sand-grain roughness $k_\text{s} = k_\text{99} (ES_x ( - ES_x + Sk + 2.37) + 0.772)$ with reasonable predictive powers. The predictive capability of the temperature roughness function is subject to limitations due to the missing Prandtl-number variation in the dataset. Nevertheless, the interpretable correlation and the neural network as well as the original dataset can be used to explore the roughness functions. The functional form of the derived correlations, along with visual analysis of these surfaces, suggests a strong relationship with roughness wavelengths, further linking them to explanations based on sheltered and windward regions.

physics.flu-dyn↗

arXiv2025

Interface reconstruction of adhering droplets for distortion correction using glare points and deep learning

The flow within adhering droplets subjected to external shear flows has a significant influence on the stability and eventual detachment of the droplets from the surface. Most commonly, the velocity field inside adhering droplets is measured by means of particle image velocimetry (PIV), which requires a correction step for distortion caused by refraction of light at the gas-liquid interface. Current methods for distortion correction based on ray tracing are limited to low external flow velocities. However, the ray-tracing method can be extended to arbitrarily deformed droplet shapes if the instantaneous three-dimensional droplet interface is availble. In the present work, a previously introduced method for the image-based reconstruction of gas-liquid interfaces by means of deep learning is adapted to determine the instantaneous interface of adhering droplets in external shear flows. In this regard, a purposefully developed optical measurement technique based on the shadowgraphy method is employed that encodes additional three-dimensional (3D) information of the interface in the images via glare points from lateral light sources. On the basis of the images recorded in the experiments, the volumetric shape of the droplet is reconstructed by a neural network that was trained on the spatio-temporal dynamics of the gas-liquid interface from a synthetic dataset obtained by numerical simulation. The results for experiments with adhering droplets at different velocities of external flow demonstrate that the combination of the learned droplet geometry with the depth encoding through the glare points facilitates a robust and flexible reconstruction. The proposed method reconstructs the instantaneous three-dimensional interface of adhering droplets at both high resolution and spatial accuracy and thereby enables the distortion correction of PIV measurements at high external flow velocities.

physics.flu-dyn↗

arXiv2024

PINNs4Drops: Video-conditioned physics-informed neural networks for two-phase flow reconstruction

Two-phase flow phenomena underpin critical technologies such as hydrogen fuel cells, spray cooling, and combustion, where droplet dynamics govern performance and efficiency. Conventional optical diagnostics, including shadowgraphy and particle image velocimetry, provide valuable insights but are limited to two-dimensional projections of inherently three-dimensional flows. We employ a specialized optical technique that encodes droplet surface information through color-coded glare points, enabling enhanced reconstruction of gas-liquid interfaces. To interpret these measurements, we introduce video-conditioned physics-informed neural networks (VcPINNs), which integrate experimental observations with governing fluid dynamics equations. This hybrid framework leverages the strengths of both data-driven learning and physical constraints, allowing accurate volumetric flow reconstruction from limited input images. Applied to droplet impingement experiments, our method yields highly resolved and physically consistent 3D interface and flow dynamics. The combined imaging and PINN reconstruction strategy provides a powerful platform for advancing multiphase-flow analysis, with broad potential impact across energy, cooling, and propulsion applications.

physics.flu-dyn↗

arXiv2024

Investigating the shortcomings of the Flow Convergence Method for quantification of Mitral Regurgitation in a pulsatile in-vitro environment and with Computational Fluid Dynamics

The flow convergence method includes calculation of the proximal isovelocity surface area (PISA) and is widely used to classify mitral regurgitation (MR) with echocardiography. It constitutes a primary decision factor for determination of treatment and should therefore be a robust quantification method. However, it is known for its tendency to underestimate MR and its dependence on user expertise. The present work systematically compares different pulsatile flow profiles arising from different regurgitation orifices using transesophageal echocardiographic (TEE) probe and particle image velocimetry (PIV) as a reference in an in-vitro environment. It is found that the inter-observer variability using echocardiography is small compared to the systematic underestimation of the regurgitation volume for large orifice areas (up to 52%) where a violation of the flow convergence method assumptions occurs. From a flow perspective, a starting vortex was found as a dominant flow pattern in the regurgant jet for all orifice shapes and sizes. A series of simplified computational fluid dynamics (CFD) simulations indicate that selecting a suboptimal aliasing velocity during echocardiography measurements might be a primary source of potential underestimation in MR characterization via the PISA-based method, reaching up to 40%. In this study, it has been noted in clinical observations that physicians often select an aliasing velocity higher than necessary for optimal estimation in diagnostic procedures.

physics.flu-dyn↗

arXiv2023

Spatio-temporal reconstruction of drop impact dynamics by means of color-coded glare points and deep learning

The present work introduces a deep learning approach for the three-dimensional reconstruction of the spatio-temporal dynamics of the gas-liquid interface in two-phase flows on the basis of monocular images obtained via optical measurement techniques. The dynamics of liquid droplets impacting onto structured solid substrates are captured through high-speed imaging in an extended shadowgraphy setup with additional reflective glare points from lateral light sources that encode further three-dimensional information of the gas-liquid interface in the images. A neural network is learned for the physically correct reconstruction of the droplet dynamics on a labelled dataset generated by synthetic image rendering on the basis of gas-liquid interface shapes obtained from direct numerical simulation. The employment of synthetic image rendering allows for the efficient generation of training data and circumvents the introduction of errors resulting from the inherent discrepancy of the droplet shapes between experiment and simulation. The accurate reconstruction of the gas-liquid interface during droplet impingement on the basis of images obtained in the experiment demonstrates the practicality of the presented approach based on neural networks and synthetic training data generation. The introduction of glare points from lateral light sources in the experiments is shown to improve the reconstruction accuracy, which indicates that the neural network learns to leverage the additional three-dimensional information encoded in the images for a more accurate depth estimation. Furthermore, the physically reasonable reconstruction of unknown gas-liquid interface shapes indicates that the neural network learned a versatile model of the involved two-phase flow phenomena during droplet impingement.

physics.flu-dyn↗

arXiv2023

Prediction of equivalent sand-grain size and identification of drag-relevant scales of roughness -- a data driven approach

The purpose of the present work is to examine two possibilities; firstly, predicting equivalent sand-grain roughness size $k_s$ based on the roughness height probability density function and power spectrum leveraging machine learning as a regression tool, and secondly, extracting information about relevance of different roughness scales to skin-friction drag by interpreting the output of the trained data-driven model. The model is an ensemble neural network consisting of 50 deep neural networks. The data for the training of the model is obtained from direct numerical simulations (DNSs) of turbulent flow in plane channels over 85 irregular multi-scale roughness samples at friction Reynolds number Re$_\tau=800$. The 85 roughness samples are selected from a repository of 4200 samples, covering a wide parameter space, through an active learning (AL) framework. The selection is made in several iterations, based on the informativeness of samples in the repository, quantified by the variance of ENN predictions. This AL framework aims to maximize the generalizability of the predictions with a certain amount of data. This is examined using three different testing data sets with different types of roughness, including 21 surfaces from the literature. The model yields an overall mean error of 5\% to 10\% on different testing data sets. Subsequently, a data interpretation technique, known as layer-wise relevance propagation, is applied to measure the contributions of different roughness wave-lengths to the predicted $k_s$. High-pass filtering is then applied to the roughness PS to exclude the wave-numbers identified as drag-irrelevant. The filtered rough surfaces are investigated using DNS, and it is demonstrated that, despite significant impact of filtering on the roughness topographical appearance and statistics, the skin-friction coefficient of the original roughness is successfully preserved.

physics.flu-dyn↗

arXiv2022

Actively Learning Costly Reward Functions for Reinforcement Learning

Transfer of recent advances in deep reinforcement learning to real-world applications is hindered by high data demands and thus low efficiency and scalability. Through independent improvements of components such as replay buffers or more stable learning algorithms, and through massively distributed systems, training time could be reduced from several days to several hours for standard benchmark tasks. However, while rewards in simulated environments are well-defined and easy to compute, reward evaluation becomes the bottleneck in many real-world environments, e.g., in molecular optimization tasks, where computationally demanding simulations or even experiments are required to evaluate states and to quantify rewards. Therefore, training might become prohibitively expensive without an extensive amount of computational resources and time. We propose to alleviate this problem by replacing costly ground-truth rewards with rewards modeled by neural networks, counteracting non-stationarity of state and reward distributions during training with an active learning component. We demonstrate that using our proposed ACRL method (Actively learning Costly rewards for Reinforcement Learning), it is possible to train agents in complex real-world environments orders of magnitudes faster. By enabling the application of reinforcement learning methods to new domains, we show that we can find interesting and non-trivial solutions to real-world optimization problems in chemistry, materials science and engineering.

cs.LG↗

arXiv2021

DNS-based characterization of pseudo-random roughness in minimal channels

Direct numerical simulation is used to study turbulent flow over irregular rough surfaces in the periodic minimal channel configuration. The generation of irregular rough surface is based on a random algorithm, in which the power spectrum of the roughness height function along with its probability density function can be directly prescribed. The hydrodynamic properties of the roughness are investigated and compared to those obtained from full-size DNS for 12 roughness topographies with systematically varied PDF and PS at four roughness height. The comparison confirms the viability of the minimal channel approach for characterization of rough surfaces providing excellent agreement in roughness function and zero-plane displacement across various types of roughness and different regimes. Results also indicates that different realizations of roughness, with a fixed PS and PDF, translate to similar values of roughness function with a small scatter. In addition to the global flow properties, the distribution of time-averaged surface force exerted by the roughness onto the fluid is examined and compared to the roughness height distribution for different cases. It is shown that the surface force distribution has an anisotropic structure with spanwise-elongated coherent regions. The anisotropy translates into a very small streamwise integral length scale, which weakly depends on the considered roughness topography, while the larger spanwise integral length scale shows a stronger dependence on roughness characteristics. It is also shown that the sheltering model describes well the spatial distribution of the surface force. Finally, existing roughness correlations are assessed using the present dataset. It was shown that the most correlations can reproduce the values of equivalent sand-grain roughness from DNS within+-30% error while none of the correlations shows a superior predictive accuracy.

physics.flu-dyn↗

arXiv2021

Predicting drag on rough surfaces by transfer learning of empirical correlations

Recent developments in neural networks have shown the potential of estimating drag on irregular rough surfaces. Nevertheless, the difficulty of obtaining a large high-fidelity dataset to train neural networks is deterring their use in practical applications. In this study, we propose a transfer learning framework to model the drag on irregular rough surfaces even with a limited amount of direct numerical simulations. We show that transfer learning of empirical correlations, reported in the literature, can significantly improve the performance of neural networks for drag prediction. This is because empirical correlations include `approximate knowledge' of the drag dependency in high-fidelity physics. The `approximate knowledge' allows neural networks to learn the surface statistics known to affect drag more efficiently. The developed framework can be applied to applications where acquiring a large dataset is difficult, but empirical correlations have been reported.

physics.flu-dyn↗

arXiv2021

Machine learning for rapid discovery of laminar flow channel wall modifications that enhance heat transfer

Numerical simulation of fluids plays an essential role in modeling many physical phenomena, which enables technological advancements, contributes to sustainable practices, and expands our understanding of various natural and engineered systems. The calculation of heat transfer in fluid flow in simple flat channels is a relatively easy task for various simulation methods. However, once the channel geometry becomes more complex, numerical simulations become a bottleneck in optimizing wall geometries. We present a combination of accurate numerical simulations of arbitrary, flat, and non-flat channels and machine learning models predicting drag coefficient and Stanton number. We show that convolutional neural networks (CNN) can accurately predict the target properties at a fraction of the time of numerical simulations. We use the CNN models in a virtual high-throughput screening approach to explore a large number of possible, randomly generated wall architectures. Data Augmentation was applied to existing geometries data to add generated new training data which have the same number of parameters of heat transfer to improve the model's generalization. The general approach is not only applicable to simple flow setups as presented here but can be extended to more complex tasks, such as multiphase or even reactive unit operations in chemical engineering.

physics.flu-dyn↗