arXiv subjects

Matthias Ihme

Publications and source records attributed to Matthias Ihme.

At least 37 records · Page 2Linked to original sources

arXiv2022

A TensorFlow Simulation Framework for Scientific Computing of Fluid Flows on Tensor Processing Units

A computational fluid dynamics (CFD) simulation framework for fluid-flow prediction is developed on the Tensor Processing Unit (TPU) platform. The TPU architecture is featured with accelerated dense matrix multiplication, large high bandwidth memory, and a fast inter-chip interconnect, making it attractive for high-performance scientific computing. The CFD framework solves the variable-density Navier-Stokes equation using a low-Mach approximation, and the governing equations are discretized by a finite-difference method on a collocated structured mesh. It uses the graph-based TensorFlow as the programming paradigm. The accuracy and performance of this framework is studied both numerically and analytically, specifically focusing on effects of TPU-native single precision floating point arithmetic. The algorithm and implementation are validated with canonical 2D and 3D Taylor-Green vortex simulations. To demonstrate the capability for simulating turbulent flows, simulations are conducted for two configurations, namely decaying homogeneous isotropic turbulence and a turbulent planar jet. Both simulations show good statistical agreement with reference solutions. The performance analysis shows a linear weak scaling and a superlinear strong scaling up to a full TPU v3 pod with 2048 cores.

physics.comp-ph↗

arXiv2021

Convolutional LSTM Neural Networks for Modeling Wildland Fire Dynamics

As the climate changes, the severity of wildland fires is expected to worsen. Models that accurately capture fire propagation dynamics greatly help efforts for understanding, responding to and mitigating the damages caused by these fires. Machine learning techniques provide a potential approach for developing such models. The objective of this study is to evaluate the feasibility of using a Convolutional Long Short-Term Memory (ConvLSTM) recurrent neural network to model the dynamics of wildland fire propagation. The machine learning model is trained on simulated wildfire data generated by a mathematical analogue model. Three simulated datasets are analyzed, each with increasing degrees of complexity. The simplest dataset includes a constant wind direction as a single confounding factor, whereas the most complex dataset includes dynamic wind, complex terrain, spatially varying moisture content and heterogenous vegetation density distributions. We examine how effective the ConvLSTM can learn the fire-spread dynamics over consecutive time steps. It is shown that ConvLSTMs can capture local fire transmission events, as well as the overall fire dynamics, such as the rate at which the fire spreads. Finally, we demonstrate that ConvLSTMs outperform other network architectures that have previously been used to model similar wildland fire dynamics.

cs.LG↗

arXiv2021

Interpretable Data-driven Methods for Subgrid-scale Closure in LES for Transcritical LOX/GCH4 Combustion

Many practical combustion systems such as those in rockets, gas turbines, and internal combustion engines operate under high pressures that surpass the thermodynamic critical limit of fuel-oxidizer mixtures. These conditions require the consideration of complex fluid behaviors that pose challenges for numerical simulations, casting doubts on the validity of existing subgrid-scale (SGS) models in large-eddy simulations of these systems. While data-driven methods have shown high accuracy as closure models in simulations of turbulent flames, these models are often criticized for lack of physical interpretability, wherein they provide answers but no insight into their underlying rationale. The objective of this study is to assess SGS stress models from conventional physics-driven approaches and an interpretable machine learning algorithm, i.e., the random forest regressor, in a turbulent transcritical non-premixed flame. To this end, direct numerical simulations (DNS) of transcritical liquid-oxygen/gaseous-methane (LOX/GCH4) inert and reacting flows are performed. Using this data, a priori analysis is performed on the Favre-filtered DNS data to examine the accuracy of physics-based and random forest SGS-models under these conditions. SGS stresses calculated with the gradient model show good agreement with the exact terms extracted from filtered DNS. The accuracy of the random-forest regressor decreased when physics-based constraints are applied to the feature set. Results demonstrate that random forests can perform as effectively as algebraic models when modeling subgrid stresses, only when trained on a sufficiently representative database. The employment of random forest feature importance score is shown to provide insight into discovering subgrid-scale stresses through sparse regression.

physics.flu-dyn↗

arXiv2021

Heat transfer augmentation by recombination reactions in turbulent reacting boundary layers at elevated pressures

A study of a reacting boundary layer flow with heat transfer at conditions typical for configurations at elevated pressures has been performed using a set of direct numerical simulations. Effects of wall temperatures are investigated, representative for cooled walls of gas turbines and sub-scale rocket engines operating with hydrocarbon as fuels. The results show that exothermic chemical reactions induced by the low-enthalpy in the boundary layer take place predominantly in the logarthimic sub-layer. The majority of the heat release is attributed to the exothermic recombination of OH and CO to produce CO2 and H2O. The recombination reactions result in an increase of the wall heat loads by up to 20% compared to the inert flow. The gas composition experiences strong deviations from the chemical equilibrium conditions. In fact, a quenching of the major species is observed within the viscous sub-layer and the transition region. Analysis of chemical time-scales shows that the location of quenched composition coincides with the region where the Damkoehler number decreases below unity. Within the viscous sub-layer, a secondary reaction zone is detected, involving the production of formyl and formaldehyde radicals that provide an additional source of energy release. The analysis of the reaction paths showed that reactions with zero activation energy are responsible for this change in gas composition, which also account for the initial branching of hydrocarbon fuels decomposition according to previous auto-ignition studies. The effect of the secondary recombination reactions is more prominent for the lower wall temperature case. Finally, the role of turbulent fluctuations on the species net chemical production rates is evaluated, showing a strong correlation between species and temperature fluctuations.

physics.flu-dyn↗

arXiv2021

Deep Learning Models for Predicting Wildfires from Historical Remote-Sensing Data

Identifying regions that have high likelihood for wildfires is a key component of land and forestry management and disaster preparedness. We create a data set by aggregating nearly a decade of remote-sensing data and historical fire records to predict wildfires. This prediction problem is framed as three machine learning tasks. Results are compared and analyzed for four different deep learning models to estimate wildfire likelihood. The results demonstrate that deep learning models can successfully identify areas of high fire likelihood using aggregated data about vegetation, weather, and topography with an AUC of 83%.

cs.CV↗

arXiv2021

Data-assisted combustion simulations with dynamic submodel assignment using random forests

In this investigation, we outline a data-assisted approach that employs random forest classifiers for local and dynamic combustion submodel assignment in turbulent-combustion simulations. This method is applied in simulations of a single-element GOX/GCH4 rocket combustor; a priori as well as a posteriori assessments are conducted to (i) evaluate the accuracy and adjustability of the classifier for targeting different quantities-of-interest (QoIs), and (ii) assess improvements, resulting from the data-assisted combustion model assignment, in predicting target QoIs during simulation runtime. Results from the a priori study show that random forests, trained with local flow properties as input variables and combustion model errors as training labels, assign three different combustion models - finite-rate chemistry (FRC), flamelet progress variable (FPV) model, and inert mixing (IM) - with reasonable classification performance even when targeting multiple QoIs. Applications in a posteriori studies demonstrate improved predictions from data-assisted simulations, in temperature and CO mass fraction, when compared with monolithic FPV calculations. These results demonstrate that this data-driven framework holds promise for the dynamic combustion submodel assignment in reacting flow simulations.

physics.flu-dyn↗

arXiv2020

Requirements Towards Predictive Simulations of Turbulent Reacting Flows

Significant progress has been made on the model development for simulating turbulent reacting flows. As a consequence, we are currently in a position where key-physical aspects of fairly complex combustion processes are well understood at a qualitative and -- in many cases -- also at a quantitative level. Examples are the prediction of temperature and major species, statistically stationary flames, gas-phase combustion, turbulent transport, and turbulence/flame coupling. However, current challenges lie in capturing transient processes and stability boundaries, minor species and emissions, multiphase flows and phase-transition, as well as multidimensional flame/flow interactions that may involve flame curvature effects, stratification, partially premixing, or flame-wall interaction. With this, the question arises what steps need to be taken to elevate the current state of modeling capabilities in order to address these deficiencies? This paper seeks to address this question. We begin by reviewing the current state of combustion model approaches, our quest for improving existing models, the separation of errors arising from numerical discretization and physical models, and ideas on model evaluations. We then proceed by examining concepts on quantitative model evaluations, requirements on predictability, quantities of interest, and cost/accuracy trade-offs. We close by introducing recent concepts that assimilate time-resolved measurements into numerical simulations for state estimation, model evaluation, and parameter determination.

physics.flu-dyn↗

arXiv2020

A General Drag Coefficient for Flow over a Sphere

A generalized physics-based expression for the drag coefficient of spherical particles moving in a fluid is derived. The proposed correlation incorporates essential rarefied physics, low-speed hydrodynamics, and shock-wave physics to accurately model the particle-drag force for a wide range of Mach and Knudsen numbers (and therefore Reynolds number) a particle may experience. Owing to the basis of the derivation in physics-based scaling laws, the proposed correlation embeds gas-specific properties and has explicit dependence on the ratio of specific heat capacities at constant pressure and constant volume. The correlation is applicable for arbitrary particle relative velocity, particle diameter, gas pressure, gas temperature, and surface temperature. Compared to existing drag models, the correlation is shown to more accurately reproduce a wide range of experimental data. Finally, the new correlation is applied to simulate dust particles' trajectories in high-speed flow, relevant to a spacecraft entering the Martian atmosphere. The enhanced surface heat flux due to particle impact is found to be sensitive to the particle drag model.

physics.flu-dyn↗

arXiv2018

Computational Modeling and Analysis of Diesel-fuel Injection and Autoignition at Transcritical Conditions

The need for improved engine efficiencies has motivated the development of high-pressure combustion systems, in which operating conditions achieve and exceed critical conditions. Associated with these conditions are strong variations in thermo-transport properties as the fluid undergoes phase transition, and two-stage ignition with low-temperature combustion. Accurately simulating these physical phenomena at real-fluid environments remains a challenge. By addressing this issue, a high-fidelity LES-modeling framework is developed to conduct simulations of transcritical fuel spray mixing and auto-ignition at high-pressure conditions. The simulation is based on a recently developed diffused interface method that solves the compressible multi-species conservation equations along with a Peng-Robinson state equation and real-fluid transport properties. LES analysis is performed for non-reacting and reacting spray conditions targeting the ECN Spray A configuration at chamber conditions with a pressure of 60 bar and temperatures between 900 K and 1200 K to investigate effects of the real-fluid environment and low-temperature chemistry. Comparisons with measurements in terms of global spray parameters (i.e., liquid and vapor penetration lengths) are shown to be in good agreement. Analysis of the mixture fraction distributions in the dispersed spray region demonstrates the accuracy in modelling the turbulent mixing behavior. Good agreement of the ignition delay time and the lift-off length is obtained from simulation results at different ambient temperature conditions and the formation of intermediate species is captured by the simulations, indicating that the presented numerical framework adequately reproduces the corresponding low- and high-temperature ignition processes under high-pressure conditions, which are relevant to realistic diesel-fuel injection systems.

physics.flu-dyn↗

arXiv2018

Transcritical Mixing and Auto-Ignition of n-dodecane Liquid Fuel using a Diffused Interface Method

High-fidelity simulations of transcritical spray mixing and auto ignition in a combustion chamber are performed at high pressure and temperature conditions using a recently developed finite rate LES solver. The simulation framework is based on a diffused-interface (DI) method that solves the compressible multi-species conservation equations along with the Peng Robinson state equation and real-fluid transport properties. A finite volume approach with entropy stable scheme is employed to accurate simulate the non-linear real fluid flow. LES analysis is performed for non-reacting and reacting spray conditions targeting the ECN Spray A configuration at chamber conditions with a pressure of 60 bar and temperatures between 800 K and 1200 K to investigate effects of the real-fluid environment and low-temperature chemistry. Comparisons with measurements in terms of global spray parameters and mixture fraction distributions demonstrates the accuracy in modeling the turbulent mixing behavior. Good overall agreement of the auto-ignition process is obtained from simulation results at different ambient temperature conditions and the formation of intermediate species is captured by the simulations, indicating that the presented numerical framework adequately reproduces the corresponding low-and-high-temperature ignition processes under high-pressure conditions that are relevant to realistic diesel fuel injection systems.

physics.flu-dyn↗

arXiv2018

Lyapunov exponent and Wasserstein metric as validation tools for assessing short-time dynamics and quantitative model evaluation of large-eddy simulation

In this work, methods for the evaluation of LES-quality and LES-accuracy are presented, which include the Lyapunov exponent for the analysis of short-time predictability of LES-calculation and the Wasserstein metric for the quantitative assessment of simulation results. Both methods are derived and evaluated in application to the Volvo test case. Both the non-reacting and reacting cases are calculated. For the non- reacting cases, good agreement with the experimental data is achieved by solvers at high numerical resolution. The reacting cases are more challenging due to the small length scale of the flame and the suppression of sinuous mode of absolute instability by the density ratio. The analysis of the turbulent simulation data using the concept of the Lyapunov exponent and the Wasserstein metric provides a more quantitative approach to assess the mesh dependency of the simulation results. The convergence of the Lyapunov exponent is shown to be a more sensitive and stronger indication of mesh-independence. Though grid convergence for the reacting cases cannot be reached with the chosen resolutions, the Lyapunov exponents and the Wasserstein metric are shown to be capable of identifying quantity-specific sensitivities with respect to the numerical resolution, while requiring significantly less computational resources than acquiring profiles of conventional turbulent statistics.

physics.flu-dyn↗

arXiv2017

Direct numerical simulations of turbulent channel flow under transcritical conditions

Turbulent flows under transcritical conditions are present in regenerative cooling systems of rocker engines and extraction processes in chemical engineering. The turbulent flows and the corresponding heat transfer phenomena in these complex processes are still not well understood experimentally and numerically. The objective of this work is to investigate the turbulent flows under transcritical conditions using DNS of turbulent channel flows. A fully compressible solver is used in conjunction with a Peng-Robinson real-fluid equation of state to describe the transcritical flows. A channel flow with two isothermal walls is simulated with one heated and one cooled boundary layers. The grid resolution adopted in this study is slightly finer than that required for DNS of incompressible channel flows. The simulations are conducted using both fully (FC) and quasi-conservative (QC) schemes to assess their performance for transcritical wall-bounded flows. The instantaneous flows and the statistics are analyzed and compared with the canonical theories. It is found that results from both FC and QC schemes qualitatively agree well with noticeable difference near the top heated wall, where spurious oscillations in velocity can be observed. Using the DNS data, we then examine the usefulness of Townsend attached eddy hypothesis in the context of flows at transcritical conditions. It is shown that the streamwise energy spectrum exhibits the inverse wavenumber scaling and that the streamwise velocity structure function follows a logarithmic scaling, thus providing support to the attached eddy model at transcritical conditions.

physics.flu-dyn↗

arXiv2017

Numerical analysis on mixing processes for transcritical real-fluid simulations

The accurate and robust simulation of transcritical real-fluid flows is crucial for many engineering applications. Diffused interface methods are frequently employed and several numerical schemes have been developed for simulating transcritical flows. These schemes can be categorized into two types, namely fully conservative and quasi-conservative schemes. An adaptive scheme which is a hybrid of the two is developed in this study. By considering several numerical test cases, it is shown that different schemes predict distinctly different mixing behaviors. It is shown that the mixing processes follow the isobaric-adiabatic and isobaric-isochoric mixing models for fully and quasi-conservative schemes, respectively, and the adaptive scheme yields a mixing behavior that spans both models. The distinct mixing behaviors are a consequence of numerical diffusion instead of physical diffusion and can be attributed to insufficient numerical spatial resolution. This work provides a better understanding on the interpretation of numerical simulation results and the mixing models that are commonly used to study transcritical flows.

physics.flu-dyn↗

arXiv2017

Efficient time stepping for reactive turbulent simulations with stiff chemistry

A combination of a steady-state preserving operator splitting method and a semi-implicit integration scheme is proposed for efficient time stepping in simulations of unsteady reacting flows, such as turbulent flames, using detailed chemical kinetic mechanisms. The operator splitting is based on the Simpler balanced splitting method, which is constructed with improved stability properties and reduced computational cost. The method is shown to be capable of stable and accurate prediction of ignition and extinction for reaction-diffusion systems near critical conditions. The ROK4E scheme is designed for semi-implicit integration of spatially independent chemically reacting systems. Being a Rosenbrock-Krylov method, ROK4E utilizes the low-rank approximation of the Jacobian to reduce the cost for integrating the system of ODEs that have relative few stiff components. The efficiency of the scheme is further improved via the careful choice of coefficients to require three right-hand-side evaluations over four stages. Combing these two methods, efficient calculation is achieved for large-scale parallel simulations of turbulent flames.

physics.comp-ph↗

arXiv2017

Application of Pareto-efficient combustion modeling framework to large eddy simulations of turbulent reacting flows

In the application of the combustion models based on low-dimensional manifolds (for instance flamelet models) to large-eddy simulation (LES) of reactive turbulent flows, the modeling simplifications of the combustion process is a critical source of uncertainty in addition to those due to the turbulent closure model and numerical discretization. The ability to quantitatively assess this uncertainty in absence of the reference result is vital to the reliable and predictive simulations of practical combustion devices.In the present study, the Pareto-efficient combustion (PEC) framework is extended to adaptive LES combustion simulations of turbulent flames. The key component of the PEC framework is the so-called manifold drift term. Its extension LES is proposed to make such assessment by examining the compliance of a particular combustion model in describing a quantity of interest with respect to the underlying flow field representation. With the focus on improving predictions of CO emissions of flamelet-based combustion models, this work employs PEC to augment the flamelet/progress variable (FPV) model through local sub-model assignment of the finite-rate chemistry (FRC) model. To this end, a series of LES-PEC calculations are performed on a piloted partially-premixed dimethyl ether flame (DME-D), using a combination of FPV and FRC models. The drift term is utilized in the PEC framework to the estimate the model related error for quantities of interest. The PEC approach is demonstrated to be capable of significantly improving the prediction of CO emissions compared with the monolithic FPV simulation. The improved accuracy is achieved by enriching the FPV model with FRC in regions where the lower-order model is determined insufficient through the evaluation of drift terms.

physics.flu-dyn↗

arXiv2017

Compositional inhomogeneities as a source of indirect combustion noise

The generation of indirect combustion noise by compositional inhomogeneities is examined theoretically. For this, the compact nozzle theory of~\cite{MARBLE_CANDEL_JSV1977} is extended to a multi-component gas mixture, and the chemical potential function is introduced as an additional acoustic source mechanism. Transfer functions for subcritical and supercritical nozzle flows are derived and the contribution of compositional noise is compared to entropy noise and direct noise by considering an idealized nozzle downstream of the combustor exit. It is shown that compositional noise is dependent on the local mixture composition and can exceed entropy noise for fuel-lean conditions and supercritical nozzle flows. This suggests that the compositional indirect noise requires potential consideration with the implementation of low-emission combustors.

physics.flu-dyn↗

arXiv2017

MVP-Workshop Contribution: Modeling of Volvo Bluff Flame Experiment and Comparison of Finite-Volume and Discontinuous-Galerkin Schemes

The Volvo burner features the canonical configuration of a bluff-body stabilized premixed flame. This configuration was studied experimentally under the Volvo Flygmotor AB program. Two cases are considered in this study: a non-reacting case with an inlet flow speed of 16.6 m/s and a reacting case with equilibrium ratio of 0.65 and inflow speed of 17.3 m/s. The characteristic vortex shedding in the wake behind the bluff body is present in the non-reacting case, while two oscillation modes are intermittently present in the reacting case. A series of large-eddy simulations are performed on this configuration using two solvers, one using a high-resolution finite-volume (FV) scheme and the other featuring a high-order discontinuous-Galerkin (DG) discretization. The FV calculations are conducted on hexahedral meshes with three different resolution (4mm, 2mm, and 1mm). The DG calculations are performed using two different polynomial orders on the same tetrahedral mesh. For the non-reacting cases, good agreement with respect to the experimental data is achieved by both solvers at high numerical resolution. The reacting cases are calculated using a two-step global mechanism in combination with the thickened-flame model. Reasonable agreement with experiments is obtained by both solvers at higher resolution. Models for combustion-turbulence interaction are necessary for the reacting case as it contains the length scale of the flame, which is smaller than the grid resolution in all calculations. The impact of such models on the flame stability and flow/flame dynamics is the subject of future research.

physics.comp-ph↗

arXiv2017

A General Probabilistic Approach for Quantitative Assessment of LES Combustion Models

The Wasserstein metric is introduced as a probabilistic method to enable quantitative evaluations of LES combustion models. The Wasserstein metric can directly be evaluated from scatter data or statistical results using probabilistic reconstruction against experimental data. The method is derived and generalized for turbulent reacting flows, and applied to validation tests involving the Sydney piloted jet flame. It is shown that the Wasserstein metric is an effective validation tool that extends to multiple scalar quantities, providing an objective and quantitative evaluation of model deficiencies and boundary conditions on the simulation accuracy. Several test cases are considered, beginning with a comparison of mixture-fraction results, and the subsequent extension to reactive scalars, including temperature and species mass fractions of \ce{CO} and \ce{CO2}. To demonstrate the versatility of the proposed method in application to multiple datasets, the Wasserstein metric is applied to a series of different simulations that were contributed to the TNF-workshop. Analysis of the results allowed to identify competing contributions to model deviations, arising from uncertainties in the boundary conditions and model deficiencies. These applications demonstrate that the Wasserstein metric constitutes an easily applicable mathematical tool that reduce multiscalar combustion data and large datasets into a scalar-valued quantitative measure.

physics.data-an↗