arXiv subjects

Nikolaus A. Adams

Publications and source records attributed to Nikolaus A. Adams.

At least 37 records · Page 2Linked to original sources

arXiv2023

A variable speed of sound formulation for weakly compressible smoothed particle hydrodynamics

We present a Weakly Compressible SPH (WCSPH) formulation with a temporally variable speed of sound. The benefits of a time-varying sound speed formulation and the weaknesses of a constant sound speed formulation are worked out. It is shown how a variable sound speed can improve the performance, accuracy, and applicability of the WCSPH method. In our novel Uniform Compressible SPH (UCSPH) method, the required artificial speed of sound is calculated at each time step based on the current flow field. The method's robustness, performance, and accuracy are demonstrated with three test cases: a Taylor-Green vortex flow, a falling droplet impact, and a dam break. For all showcases, we observe at least similar accuracy as computed with WCSPH at strongly improved computational performance.

physics.comp-ph↗

arXiv2023

Extended Eulerian SPH and its realization of FVM

Eulerian smoothed particle hydrodynamics (Eulerian SPH) is considered as a potential meshless alternative to a traditional Eulerian mesh-based method, i.e. finite volume method (FVM), in computational fluid dynamics (CFD). While researchers have analyzed the differences between these two methods, a rigorous comparison of their performance and computational efficiency is hindered by the constraint related to the normal direction of interfaces in pairwise particle interactions within Eulerian SPH framework. To address this constraint and improve numerical accuracy, we introduce Eulerian SPH extensions, including particle relaxation to satisfy zero-order consistency, kernel correction matrix to ensure first-order consistency and release the constraint associated with the normal direction of interfaces, as well as dissipation limiters to enhance numerical accuracy and these extensions make Eulerian SPH rigorously equivalent to FVM. Furthermore, we implement mesh-based FVM within SPHinXsys, an open-source SPH library, through developing a parser to extract necessary information from the mesh file which is exported in the MESH format using the commercial software ICEM. Therefore, these comprehensive approaches enable a rigorous comparison between these two methods.

physics.flu-dyn↗

arXiv2023

Learning Lagrangian Fluid Mechanics with E($3$)-Equivariant Graph Neural Networks

We contribute to the vastly growing field of machine learning for engineering systems by demonstrating that equivariant graph neural networks have the potential to learn more accurate dynamic-interaction models than their non-equivariant counterparts. We benchmark two well-studied fluid-flow systems, namely 3D decaying Taylor-Green vortex and 3D reverse Poiseuille flow, and evaluate the models based on different performance measures, such as kinetic energy or Sinkhorn distance. In addition, we investigate different embedding methods of physical-information histories for equivariant models. We find that while currently being rather slow to train and evaluate, equivariant models with our proposed history embeddings learn more accurate physical interactions.

cs.LG↗

arXiv2023

A 2D hybrid method for interfacial transport of passive scalars

A hybrid Eulerian-Lagrangian method is proposed to simulate passive scalar transport on arbitrary shape interface. In this method, interface deformation is tracked by an Eulerian method while the transport of the passive scalar on the material interface is solved by a single-layer Lagrangian particle method. To avoid particle clustering, a novel remeshing approach is proposed. This remeshing method can resample particles, adjust the position of particles by a relaxation process, and transfer mass from pre-existing particles to resampled particles via a redistribution process, which preserves mass both globally and locally. Computational costs are controlled by an adaptive remeshing strategy. Accuracy is assessed by a series of test cases.

physics.comp-ph↗

arXiv2023

An efficient and robust all-Mach consistent numerical scheme for simulation of compressible multi-component fluids including surface tension, cavitation, turbulence modeling and interface sharpening on compact stencils

We present an efficient, fully conservative numerical scheme valid in the entire range of highly to weakly compressible flows using a single-fluid four equation approach together with multi-component thermodynamic models. The approach can easily be included into existing finite volume methods on compact stencils and enables handling of compressibility of all involved phases including surface tension, cavitation and viscous effects. The mass fraction (indicator function) is sharpened in the two-phase interface region using the algebraic interface sharpening technique Tangent of Hyperbola for INterface Capturing (THINC). The cell face reconstruction procedure for mass fractions switches between an upwind-biased and a THINC-based scheme, along with proper slope limiters and a suitable compression coefficient, respectively. For additional sub-grid turbulence modeling, a fourth order central scheme is included into the switching process, along with a modified discontinuity sensor. The proposed All-Mach Riemann solver consistently merges the thermodynamic relationship of the components into the reconstructed thermodynamic variables (like density, internal energy), wherefore we call them All Mach THINC-based Thermodynamic-Dependent Update (All-Mach THINC-TDU) method. Both, liquid-gas and liquid-vapor interfaces can be sharpened. Surface tension effects are taken into account by using a Continuum Surface Force (CSF) model. An explicit, four stage low storage Runge Kutta method is used for time integration. The proposed methodology is validated against a series of reference cases, such as bubble oscillation,advection,deformation, shock-bubble interaction, a vapor bubble collapse and a multi-component shear flow. Finally, the approach is applied to simulate the three-dimensional primary break-up of a turbulent diesel jet.

physics.flu-dyn↗

arXiv2023

On the Relationships between Graph Neural Networks for the Simulation of Physical Systems and Classical Numerical Methods

Recent developments in Machine Learning approaches for modelling physical systems have begun to mirror the past development of numerical methods in the computational sciences. In this survey, we begin by providing an example of this with the parallels between the development trajectories of graph neural network acceleration for physical simulations and particle-based approaches. We then give an overview of simulation approaches, which have not yet found their way into state-of-the-art Machine Learning methods and hold the potential to make Machine Learning approaches more accurate and more efficient. We conclude by presenting an outlook on the potential of these approaches for making Machine Learning models for science more efficient.

cs.LG↗

arXiv2023

E($3$) Equivariant Graph Neural Networks for Particle-Based Fluid Mechanics

We contribute to the vastly growing field of machine learning for engineering systems by demonstrating that equivariant graph neural networks have the potential to learn more accurate dynamic-interaction models than their non-equivariant counterparts. We benchmark two well-studied fluid flow systems, namely the 3D decaying Taylor-Green vortex and the 3D reverse Poiseuille flow, and compare equivariant graph neural networks to their non-equivariant counterparts on different performance measures, such as kinetic energy or Sinkhorn distance. Such measures are typically used in engineering to validate numerical solvers. Our main findings are that while being rather slow to train and evaluate, equivariant models learn more physically accurate interactions. This indicates opportunities for future work towards coarse-grained models for turbulent flows, and generalization across system dynamics and parameters.

cs.LG↗

arXiv2022

A six-point neuron-based ENO (NENO6) scheme for compressible fluid dynamics

In this work, we introduce a deep artificial neural network (ANN) that can detect locations of discontinuity and build a six-point ENO-type scheme based on a set of smooth and discontinuous training data. While a set of candidate stencils of incremental width is constructed, the ANN instead of a classical smoothness indicator is deployed for an ENO-like sub-stencil selection. A convex combination of the candidate fluxes with the re-normalized linear weights forms the six-point neuron-based ENO (NENO6) scheme. The present methodology is inspired by the work [Fu et al., Journal of Computational Physics 305 (2016): 333-359] where contributions of candidate stencils containing discontinuities are removed from the final reconstruction stencil. The binary candidate stencil classification is performed by a well-trained ANN with high fidelity. The proposed framework shows an improved generality and robustness compared with other ANN-based schemes. The generality and performance of the proposed NENO6 scheme are demonstrated by examining one- and two-dimensional benchmark cases with different governing conservation laws and comparing to those of WENO-CU6 and TENO6-opt schemes.

physics.comp-ph↗

arXiv2022

JAX-FLUIDS: A fully-differentiable high-order computational fluid dynamics solver for compressible two-phase flows

Physical systems are governed by partial differential equations (PDEs). The Navier-Stokes equations describe fluid flows and are representative of nonlinear physical systems with complex spatio-temporal interactions. Fluid flows are omnipresent in nature and engineering applications, and their accurate simulation is essential for providing insights into these processes. While PDEs are typically solved with numerical methods, the recent success of machine learning (ML) has shown that ML methods can provide novel avenues of finding solutions to PDEs. ML is becoming more and more present in computational fluid dynamics (CFD). However, up to this date, there does not exist a general-purpose ML-CFD package which provides 1) powerful state-of-the-art numerical methods, 2) seamless hybridization of ML with CFD, and 3) automatic differentiation (AD) capabilities. AD in particular is essential to ML-CFD research as it provides gradient information and enables optimization of preexisting and novel CFD models. In this work, we propose JAX-FLUIDS: a comprehensive fully-differentiable CFD Python solver for compressible two-phase flows. JAX-FLUIDS allows the simulation of complex fluid dynamics with phenomena like three-dimensional turbulence, compressibility effects, and two-phase flows. Written entirely in JAX, it is straightforward to include existing ML models into the proposed framework. Furthermore, JAX-FLUIDS enables end-to-end optimization. I.e., ML models can be optimized with gradients that are backpropagated through the entire CFD algorithm, and therefore contain not only information of the underlying PDE but also of the applied numerical methods. We believe that a Python package like JAX-FLUIDS is crucial to facilitate research at the intersection of ML and CFD and may pave the way for an era of differentiable fluid dynamics.

physics.flu-dyn↗

arXiv2022

Numerical prediction of erosion due to a cavitating jet

We numerically investigate the erosion potential of a cavitating liquid jet by means of high-resolution finite volume simulations. As thermodynamic model, we employ a barotropic equilibrium cavitation approach, embedded into a homogeneous mixture model. To resolve the effects of collapsing vapor structures and to estimate the erosion potential, full compressibility is considered. Two different operating points featuring different cavitation intensities are investigated and their erosion potential is estimated and compared. Different methods are used for this purpose, including collapse detection (Mihatsch et al., 2015), maximum pressure distribution on the wall, and a new method of generating numerical pit equivalents. The data of numerical pit equivalents is analyzed in detail and compared with experimental data from the literature. Furthermore, a comprehensive grid study for both operating points is presented.

physics.flu-dyn↗

arXiv2021

A fully-differentiable compressible high-order computational fluid dynamics solver

Fluid flows are omnipresent in nature and engineering disciplines. The reliable computation of fluids has been a long-lasting challenge due to nonlinear interactions over multiple spatio-temporal scales. The compressible Navier-Stokes equations govern compressible flows and allow for complex phenomena like turbulence and shocks. Despite tremendous progress in hardware and software, capturing the smallest length-scales in fluid flows still introduces prohibitive computational cost for real-life applications. We are currently witnessing a paradigm shift towards machine learning supported design of numerical schemes as a means to tackle aforementioned problem. While prior work has explored differentiable algorithms for one- or two-dimensional incompressible fluid flows, we present a fully-differentiable three-dimensional framework for the computation of compressible fluid flows using high-order state-of-the-art numerical methods. Firstly, we demonstrate the efficiency of our solver by computing classical two- and three-dimensional test cases, including strong shocks and transition to turbulence. Secondly, and more importantly, our framework allows for end-to-end optimization to improve existing numerical schemes inside computational fluid dynamics algorithms. In particular, we are using neural networks to substitute a conventional numerical flux function.

physics.flu-dyn↗

arXiv2021

Numerical investigation of non-condensable gas effect on vapor bubble collapse

We numerically investigate the effect of non-condensable gas inside a vapor bubble on bubble dynamics, collapse pressure and pressure impact of spherical and aspherical bubble collapses. Free gas inside a vapor bubble has a damping effect that can weaken the pressure wave and enhance the bubble rebound. To estimate this effect numerically, we derive and validate a multi-component model for vapor bubbles containing gas. For the cavitating liquid and the non-condensable gas, we employ a homogeneous mixture model with a coupled equation of state for all components. The cavitation model for the cavitating liquid is a barotropic thermodynamic equilibrium model. Compressibility of all phases is considered in order to capture the shock wave of the bubble collapse. After validating the model with an analytical energy partitioning model, simulations of collapsing wall-attached bubbles with different stand-off distances are performed. The effect of the non-condensable gas on rebound and damping of the emitted shock wave is well captured.

physics.flu-dyn↗

arXiv2021

Mesoscopic Lattice Boltzmann Modeling of the Liquid-Vapor Phase Transition

We develop a mesoscopic lattice Boltzmann model for liquid-vapor phase transition by handling the microscopic molecular interaction. The short-range molecular interaction is incorporated by recovering an equation of state for dense gases, and the long-range molecular interaction is mimicked by introducing a pairwise interaction force. Double distribution functions are employed, with the density distribution function for the mass and momentum conservation laws and an innovative total kinetic energy distribution function for the energy conservation law. The recovered mesomacroscopic governing equations are fully consistent with kinetic theory, and thermodynamic consistency is naturally satisfied.

physics.flu-dyn↗

arXiv2021

Effect of stand-off distance and spatial resolution on the pressure impact of near-wall vapor bubble collapses

We consider the collapse behavior of cavitation bubbles near walls under high ambient pressure conditions. Generic configurations with different stand-off distances are investigated by numerical simulation using a fully compressible two-phase flow solver including phase change. The results show that the stand-off distance has significant effects on collapse dynamics, micro-jet formation, rebound, and maximum wall pressure. A relation between cavitation induced material damage and corresponding collapse mechanisms is obtained from pressure-impact data at the wall. We analyze the resolution dependence of collapse and rebound and the observed maximum pressure distributions. The comparison of the results on six different grid resolutions shows that main collapse features are already captured on the coarsest resolution, while the peak pressures are strongly resolution dependent.

physics.flu-dyn↗

arXiv2021

Large eddy simulations of cavitating flow in a step nozzle with injection into gas

We present results of large eddy simulations of a cavitating nozzle flow and injection into gas, investigating the interactions of cavitation in the nozzle, primary jet breakup, mass-flow rates, and gas entrainment. During strong cavitation, detached vapor structures can reach the nozzle outlet, leading to partial entrainment of gas from the outflow region into the nozzle. The gas entrainment can affect cavitation dynamics, mass-flow rates, and jet breakup. Moreover, the implosion of detached vapor structures induces pressure peaks that on the one hand amplify turbulent fluctuations and subsequently can enhance jet breakup and on the other hand can damage walls in the proximity and thus lead to cavitation erosion. Our numerical setup is based on a reference experiment, in which liquid water is discharged into ambient air through a step nozzle. The cavitating liquid and the non-condensable gas phase are modeled with a barotropic homogeneous mixture model while for the numerical model a high-order implicit large eddy approach is employed. Full compressibility of all components is taken into account, enabling us to capture the effects of collapsing vapor structures. Two operating points covering different cavitation regimes and jet characteristics are investigated. Special emphasis is placed on studying the effects of cavitation on the mass flow and the jet as well as the impact of partial gas entrainment. Therefore, frequency analyses of the recorded time-resolved signals are performed. Furthermore, the dynamics and intensities of imploding vapor structures are assessed.

physics.flu-dyn↗

arXiv2021

Inferring incompressible two-phase flow fields from the interface motion using physics-informed neural networks

In this work, physics-informed neural networks are applied to incompressible two-phase flow problems. We investigate the forward problem, where the governing equations are solved from initial and boundary conditions, as well as the inverse problem, where continuous velocity and pressure fields are inferred from scattered-time data on the interface position. We employ a volume of fluid approach, i.e. the auxiliary variable here is the volume fraction of the fluids within each phase. For the forward problem, we solve the two-phase Couette and Poiseuille flow. For the inverse problem, three classical test cases for two-phase modeling are investigated: (i) drop in a shear flow, (ii) oscillating drop and (iii) rising bubble. Data of the interface position over time is generated by numerical simulation. An effective way to distribute spatial training points to fit the interface, i.e. the volume fraction field, and the residual points is proposed. Furthermore, we show that appropriate weighting of losses associated with the residual of the partial differential equations is crucial for successful training. The benefit of using adaptive activation functions is evaluated for both the forward and inverse problem.

physics.flu-dyn↗

arXiv2021

Normalizing Flows as a Novel PDF Turbulence Model

In this paper, we propose normalizing flows (NF) as a novel probability density function (PDF) turbulence model (NF-PDF model) for the Reynolds-averaged Navier-Stokes (RANS) equations. We propose to use normalizing flows in two different ways: firstly, as a direct model for the Reynolds stress tensor, and secondly as a second-moment closure model, i.e. for modeling the pressure-strain and dissipation tensor in the Reynolds stress transport equation. In classical PDF closure models, a stochastic differential equation has to be modeled and solved to obtain samples of turbulent quantities. The NF-PDF closure model allows for direct sampling from the underlying probability density functions of fluctuating turbulent quantities, such that ensemble-averaged quantities can then be computed. To illustrate this approach we demonstrate an application for the canonical case of homogeneous shear turbulence. Training data are extracted from high-fidelity direct numerical simulations (DNS).

physics.flu-dyn↗

arXiv2020

A modular massively parallel computing environment for three-dimensional multiresolution simulations of compressible flows

Numerical investigation of compressible flows faces two main challenges. In order to accurately describe the flow characteristics, high-resolution nonlinear numerical schemes are needed to capture discontinuities and resolve wide convective, acoustic and interfacial scale ranges. The simulation of realistic 3D problems with state-of-the-art FVM based on approximate Riemann solvers with weighted nonlinear reconstruction schemes requires the usage of HPC architectures. Efficient compression algorithms reduce computational and memory load. Fully adaptive MR algorithms with LTS have proven their potential for such applications. While modern CPU require multiple levels of parallelism to achieve peak performance, the fine grained MR mesh adaptivity results in challenging compute/communication patterns. Moreover, LTS incur for strong data dependencies which challenge a parallelization strategy. We address these challenges with a block-based MR algorithm, where arbitrary cuts in the underlying octree are possible. This allows for a parallelization on distributed-memory machines via the MPI. We obtain neighbor relations by a simple bit-logic in a modified Morton Order. The block-based concept allows for a modular setup of the source code framework in which the building blocks of the algorithm, such as the choice of the Riemann solver or the reconstruction stencil, are interchangeable without loss of parallel performance. We present the capabilities of the modular framework with a range of test cases and scaling analysis with effective resolutions beyond one billion cells using $\mathcal{O}(10^3)$ cores.

physics.comp-ph↗