19 results
Filters
Settings
Now showing1 - 10 of 19
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Conditional neural field-based reduced order model for dynamic ditching load prediction(American Institute of Physics (AIP), 2026-10-22); ; ; Grid-based neural networks such as convolutional autoencoders are widely used in dimension reduction-based surrogate models for computational fluid dynamics. Recently, the use of coordinate-based approaches like conditional neural fields (CNFs) has emerged. Their independence of the spatial discretization is an advantageous feature and a key motivation for computational fluid dynamic applications. This paper discusses the spatiotemporal prediction of aircraft ditching loads using a CNF approach. The model employs a Fourier feature mapping of input coordinates to capture high frequency content. Based on feature-wise linear modulation, the output is conditioned on a latent vector, which is learned by autodecoding. The model is evaluated using two datasets for the dynamic loads on the fuselage of a German Aerospace Center (DLR) D150 aircraft. The first dataset relates to a single fixed spatial discretization, and the second dataset includes data from different discretizations. On the first dataset, the model is compared to a grid-dependent convolutional autoencoder-based model, and both approaches achieve similar reconstruction accuracy, but the CNF requires significantly fewer parameters. When paired with a long short-term memory network in the latent space, both approaches achieve similar spatiotemporal prediction accuracy. Results for the second dataset demonstrate the ability of the CNF-based approach to reconstruct ditching loads accurately for heterogeneous spatial discretizations. In particular, the model can generalize to discretizations not used for training. This allows for flexible use of training datasets generated for different geometries and/or discretizations and the use of the surrogate model to predict loads for different configurations.Publicationtype: Journal Article16 5 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Finite-time scalar retention and slow mixing tails in uncertain passive-scalar transportMixing of a passive scalar in a time-dependent flow is usually judged from the variance decay of a nominal flow. In many applications, though, stirring protocols and flow parameters are uncertain, and the more relevant question is whether the required level of homogenization is reached within a finite time. We study this reliability question for incompressible advection–diffusion with uncertain forcing parameters, using a time-periodic double-gyre flow as the main benchmark and a periodically stirred cavity as a secondary test. Direct numerical simulations show that delayed mixing is not a fluctuation around the mean variance decay curve. Instead, slow realizations retain coherent scalar remnants near weakly exchanging regions of the flow, which produces a persistent upper tail in the mixing time distribution. To identify these slow events without running a full simulation for every parameter sample, we build a parameter- and phase-dependent reduced transport operator by combining a proper orthogonal decomposition-based Koopman representation with a polynomial-chaos expansion in the uncertain parameters. From this operator, we compute a finite-time scalar-retention score based on the leading singular direction of the learned propagator in a scalar-energy metric. The diagnostic ranks parameter realizations by their tendency to retain scalar energy, identifies the realizations contributing to the upper tail of the mixing time distribution, and recovers the scalar structures responsible for delayed homogenization. Reliable mixing under uncertainty is therefore controlled not only by the mean variance decay, but also by the least-damped scalar directions of the uncertain transport operator, which suggests that designs targeting reliable mixing should act on these directions directly.Publicationtype: Journal Article8 16 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Bridging ocean wave physics and deep learning: physics-informed neural operators for nonlinear wavefield reconstruction in real-time(American Institute of Physics (AIP), 2025-10-07); ; Accurate real-time reconstruction of phase-resolved ocean wave fields remains a critical yet largely unsolved problem, primarily due to the absence of practical data assimilation methods for obtaining initial conditions from sparse or indirect wave measurements. While recent advances in supervised deep learning have shown potential for this purpose, they require large labeled datasets of ground truth wave data, which are infeasible to obtain in real-world scenarios. To overcome this limitation, we propose a physics-informed neural operator (PINO) framework for reconstructing spatially and temporally phase-resolved, nonlinear ocean wave fields from sparse measurements, without the need for ground truth data during training. This is achieved by embedding residuals of the free surface boundary conditions of ocean gravity waves into the loss function, constraining the solution space in a soft manner. In the current implementation, the framework is demonstrated for long-crested, unidirectional wave surfaces, where the wave propagation direction is aligned with the radar scanning direction. Within this setting, we validate our approach using highly realistic synthetic wave measurements by demonstrating the accurate reconstruction of nonlinear wave fields from both buoy time series and radar snapshots. Our results indicate that PINOs enable accurate, real-time reconstruction and generalize robustly across a wide range of wave conditions, thereby paving the way for future extensions of this framework toward multidirectional sea states and thus operational wave reconstruction in realistic marine environments.Publicationtype: Journal Article28 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Explicit and approximate solutions for a classical hyperbolic fragmentation equation using a hybrid projected differential transform method(American Institute of Physics (AIP), 2024-09-01); ; ; Singh, SukhjitPopulation balance equations are widely used to study the evolution of aerosols, colloids, liquid–liquid dispersion, raindrop fragmentation, and pharmaceutical granulation. However, these equations are difficult to solve due to the complexity of the kernel structures and initial conditions. The hyperbolic fragmentation equation, in particular, is further complicated by the inclusion of double integrals. These challenges hinder the analytical solutions of number density functions for basic kernel classes with exponential initial distributions. To address these issues, this study introduces a new approach combining the projected differential transform method with Laplace transform and Padé approximants to solve the hyperbolic fragmentation equation. This method aims to provide accurate and efficient explicit solutions to this challenging problem. The approach's applicability is demonstrated through rigorous mathematical derivation and convergence analysis using the Banach contraction principle. Additionally, several numerical examples illustrate the accuracy and robustness of this new method. For the first time, new analytical solutions for number density functions are presented for various fragmentation kernels with gamma and other initial distributions. This method significantly enhances solution quality over extended periods using fewer terms in the truncated series. The solutions are compared and verified against the finite volume method and the homotopy perturbation method, showing that the coupled approach not only estimates number density functions accurately but also captures integral moments with high precision. This research advances computational methods for particle breakage phenomena, offering potential applications in various industrial processes and scientific disciplines.Publicationtype: Journal Article9 9 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Impact of reduced gravity on food mixing and emptying in human stomach : a numerical simulation study(American Institute of Physics (AIP), 2024-06-14); 68 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Viscosities of inhomogeneous systems from generalized entropy scaling(American Institute of Physics (AIP), 2024-04-01); ; ; ; This study extends entropy scaling to inhomogeneous fluids by using the classical density functional theory together with a new viscosity reference that takes into account the influence of solid-fluid interactions on the fluid viscosity. The density functional theory uses a Helmholtz energy functional based on the perturbed-chain statistical associating fluid theory; the local residual entropy per particle is determined from the temperature derivative of the Helmholtz energy functional in combination with an appropriate weighted density profile. The weighted density calculation requires a single transferable parameter, which is adjusted to a reference molecular dynamics simulation. In particular, local viscosity values for fluids under nanoconfinement near solid-fluid interfaces are predicted using the same entropy scaling parameters as for homogeneous fluids. We validate the model by comparing viscosity and velocity profiles with results from non-equilibrium molecular dynamics simulations of a Couette flow in a slit pore. Good agreement is found between the entropy scaling model and the non-equilibrium molecular dynamics results for both the viscosity and velocity profiles of the Lennard-Jones truncated and shifted fluid. The proposed model extrapolates well to systems with different temperatures, fluid densities, and shear forces as well as to systems with different wetting behaviors. These results demonstrate that entropy scaling can be generalized to inhomogeneous fluids using an appropriate combination of residual entropy profile and viscosity reference.Publicationtype: Journal Article16 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , A small-eddy-dissipation mechanism for turbulence modeling and application to wall-bounded flowsA small-eddy-dissipation (SED) mechanism is proposed in the present study for the development of turbulence models. According to the SED mechanism, a turbulence model introduces artificial dissipation to filter out small eddies so that a lower resolution mesh can be used in the simulation. In addition, the artificial dissipation should be applied outside the energy-containing range so that the large-scale motions are not affected. A small-eddy-dissipation mixing length (SED-ML) model is developed based on the SED mechanism to calculate wall-bounded flows. A local Reynolds number y s + = | det ( ∇ u ) | 1 / 3 / s is introduced in the SED-ML model to distinguish laminar flows from turbulent flows. Therefore, in addition to fully turbulent flows, the SED-ML model can also be used to calculate weakly turbulent or laminar flows. To demonstrate the performance of the SED-ML model, turbulent channel flows with the Reynolds numbers R e τ up to 4200 are simulated. The numerical results are extensively compared with the large eddy simulation (LES) results using the classical subgrid-scale models. The numerical results show that the SED-ML model predicts the statistical results with a good accuracy, while requiring a lower mesh resolution than the classical LES models. The accuracy of the calculated statistical results can be further improved by the parameter extension. The friction coefficient f for channel flows can be extended directly from the reference solution according to the relation ∂ f / ∂ ϕ / f = − 9 , where ϕ is a dissipative strength coefficient. The energy and dissipation spectra confirm that the SED-ML model introduces significant dissipation at high wavenumbers. This feature is beneficial in maintaining the high accuracy of the simulation results and reducing the computational cost. The numerical study also shows that a sharp filtering of the turbulent kinetic energy in the wavenumber space is essential for the development of the turbulence model with high performance.Publicationtype: Journal Article35 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Two-phase flow simulations of surface waves in wind-forced conditions(American Institute of Physics (AIP), 2023-07-01); ; ; Scopus© Citations 2 94 138 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Scale-resolving turbulent channel flow simulations using a dynamic cumulant lattice Boltzmann methodThis paper reports the assessment of a modified cumulant lattice Boltzmann method in turbulent channel flows. The suggested approach is of interest for its resolution-spanning capabilities, which are scrutinized for grid resolutions between 4 and 430 wall units. With the emphasis on a recently published parameterized cumulant collision operator, we suggest a resolution-sensitive regularization supplemented by a dynamic subgrid-scale model based upon third-order cumulant expressions, which seamlessly vanishes for direct numerical simulations. The strategy is optionally supplemented by a wall function approach to adjust the wall treatment. The analysis involves the mean flow field, resolved and modeled second moments, two-point correlations, and spectral data and correlation lengths obtained for four friction Reynolds numbers (180, 550, 2000, and 5200). Results display a resolution-independent capability to simulate the attached turbulent shear flow.Publicationtype: Journal ArticleScopus© Citations 4 77 748 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Effects of end-walls on flows in a highly loaded compressor cascade with double-circular-arc bladesA numerical study is carried out to understand the flows in a highly loaded compressor cascade made of double-circular-arc blades, which were measured by Zierke and Deutsch in the late 1980s. A two-dimensional (2D) cascade with periodic boundary conditions in both pitch-wise and span-wise directions and a three-dimensional (3D) cascade with two end-walls that are far away from each other are accounted for in the study. For the incidence angle α = - 8.5 °, the numerical results of the 2D-cascade flow are in excellent accordance with the experimental data. This not only validates the numerical method used in the study but also suggests that a 2D and periodic flow was successfully generated in the experiment for this incidence angle. However, the numerical results of 2D-cascade flows for α = - 1.5 ° and 5 ° deviate from the experiment considerably because the strong effects of the end-walls on the wake are neglected in the simulation. By contrast, the simulation of 3D-cascade flows predicts an accurate pressure coefficient at the blade surface, the pressure increase coefficient, and the total pressure loss coefficient for all three incidence angles. This means that, to generate experimental data for validating numerical simulation, it is important to consider the effect of end-walls when the incidence angle is large. The numerical results also show that, for 2D-cascade flows with a low inlet turbulence intensity, the laminar-turbulent transition on the pressure surface is determined by the interaction of the Klebanoff distortions and T-S waves. The Klebanoff distortions are also clearly identified on the suction surface for α = - 8.5 °. The end-walls induce span-wise elongated disturbances, which suppress the stream-wise disturbances. The transition in 3D-cascade flows generally follows the mechanism of natural transition.Publicationtype: Journal ArticleScopus© Citations 4 52 438