Energies

Research

15 pages, 6484 KiB

Open AccessArticle

Acceleration of Chemical Kinetics Computation with the Learned Intelligent Tabulation (LIT) Method

by Majid Haghshenas, Peetak Mitra, Niccolò Dal Santo and David P. Schmidt

Energies 2021, 14(23), 7851; https://doi.org/10.3390/en14237851 - 23 Nov 2021

Cited by 10 | Viewed by 2080

In this work, a data-driven methodology for modeling combustion kinetics, Learned Intelligent Tabulation (LIT), is presented. LIT aims to accelerate the tabulation of combustion mechanisms via machine learning algorithms such as Deep Neural Networks (DNNs). The high-dimensional composition space is sampled from high-fidelity [...] Read more.

In this work, a data-driven methodology for modeling combustion kinetics, Learned Intelligent Tabulation (LIT), is presented. LIT aims to accelerate the tabulation of combustion mechanisms via machine learning algorithms such as Deep Neural Networks (DNNs). The high-dimensional composition space is sampled from high-fidelity simulations covering a wide range of initial conditions to train these DNNs. The input data are clustered into subspaces, while each subspace is trained with a DNN regression model targeted to a particular part of the high-dimensional composition space. This localized approach has proven to be more tractable than having a global ANN regression model, which fails to generalize across various composition spaces. The clustering is performed using an unsupervised method, Self-Organizing Map (SOM), which automatically subdivides the space. A dense network comprised of fully connected layers is considered for the regression model, while the network hyper parameters are optimized using Bayesian optimization. A nonlinear transformation of the parameters is used to improve sensitivity to minor species and enhance the prediction of ignition delay. The LIT method is employed to model the chemistry kinetics of zero-dimensional

H_{2} - O_{2}

and

{CH}_{4}

-air combustion. The data-driven method achieves good agreement with the benchmark method while being cheaper in terms of computational cost. LIT is naturally extensible to different combustion models such as flamelet and PDF transport models. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Graphical abstract

21 pages, 1168 KiB

Open AccessArticle

New Combustion Modelling Approach for Methane-Hydrogen Fueled Engines Using Machine Learning and Engine Virtualization

by Santiago Molina, Ricardo Novella, Josep Gomez-Soriano and Miguel Olcina-Girona

Energies 2021, 14(20), 6732; https://doi.org/10.3390/en14206732 - 16 Oct 2021

Cited by 13 | Viewed by 3232

Abstract

The achievement of a carbon-free emissions economy is one of the main goals to reduce climate change and its negative effects. Scientists and technological improvements have followed this trend, improving efficiency, and reducing carbon and other compounds that foment climate change. Since the [...] Read more.

The achievement of a carbon-free emissions economy is one of the main goals to reduce climate change and its negative effects. Scientists and technological improvements have followed this trend, improving efficiency, and reducing carbon and other compounds that foment climate change. Since the main contributor of these emissions is transportation, detaching this sector from fossil fuels is a necessary step towards an environmentally friendly future. Therefore, an evaluation of alternative fuels will be needed to find a suitable replacement for traditional fossil-based fuels. In this scenario, hydrogen appears as a possible solution. However, the existence of the drawbacks associated with the application of H₂-ICE redirects the solution to dual-fuel strategies, which consist of mixing different fuels, to reduce negative aspects of their separate use while enhancing the benefits. In this work, a new combustion modelling approach based on machine learning (ML) modeling is proposed for predicting the burning rate of different mixtures of methane (

{CH}_{4}

) and hydrogen (

H_{2}

). Laminar flame speed calculations have been performed to train the ML model, finding a faster way to obtain good results in comparison with actual models applied to SI engines in the virtual engine model framework. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

16 pages, 5072 KiB

Open AccessArticle

A Mapping Approach for Efficient CFD Simulation of Low-Speed Large-Bore Marine Engine with Pre-Chamber and Dual-Fuel Operation

by Ying Ye, Zongyu Yue, Hu Wang, Haifeng Liu, Chaohui Wu and Mingfa Yao

Energies 2021, 14(19), 6126; https://doi.org/10.3390/en14196126 - 26 Sep 2021

Cited by 4 | Viewed by 2206

Abstract

A natural-gas-diesel dual-fuel marine engine with a pre-chamber is a promising solution for ocean transportation to meet the International Maritime Organization (IMO) emission regulations. This engine system employs a pre-chamber with direct injection of diesel to ignite premixed natural gas due to its [...] Read more.

A natural-gas-diesel dual-fuel marine engine with a pre-chamber is a promising solution for ocean transportation to meet the International Maritime Organization (IMO) emission regulations. This engine system employs a pre-chamber with direct injection of diesel to ignite premixed natural gas due to its higher ignition energy, which can enable lower lean limit and higher thermal efficiency. The dual-fuel pre-chamber marine engine presents complex multi-regime combustion characteristics in- and outside the pre-chamber, thus posing challenges in its numerical simulation in a cost-effective manner. Therefore, this paper presents a three-dimensional modeling study for the multi-regime combustion in a large-bore two-stroke marine dual-fuel engine, proposing a novel mapping approach, which couples the well-stirred reactor (WSR) model with the G-equation model to achieve high computational accuracy and efficiency simultaneously. In-depth analysis is performed using representative exothermic reaction (RXR) analysis and premixed turbulent combustion fundamentals to better understand the combustion process and to provide guidance in the selection of mapping timing. The results show that the use of mapping to switch from the WSR to the G-equation model can effectively reduce the runtime significantly by 71.5%, meanwhile maintaining similar accuracies in predictions of in-cylinder pressure traces, HRR and NOx emissions, compared to using WSR all along. Additionally, the choice of mapping timing based on several parameters is preliminarily discussed. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

23 pages, 2880 KiB

Open AccessArticle

Generalization Capability of Convolutional Neural Networks for Progress Variable Variance and Reaction Rate Subgrid-Scale Modeling

by Victor Xing, Corentin Lapeyre, Thomas Jaravel and Thierry Poinsot

Energies 2021, 14(16), 5096; https://doi.org/10.3390/en14165096 - 18 Aug 2021

Cited by 8 | Viewed by 2404

Abstract

Deep learning has recently emerged as a successful approach to produce accurate subgrid-scale (SGS) models for Large Eddy Simulations (LES) in combustion. However, the ability of these models to generalize to configurations far from their training distribution is still mainly unexplored, thus impeding [...] Read more.

Deep learning has recently emerged as a successful approach to produce accurate subgrid-scale (SGS) models for Large Eddy Simulations (LES) in combustion. However, the ability of these models to generalize to configurations far from their training distribution is still mainly unexplored, thus impeding their application to practical configurations. In this work, a convolutional neural network (CNN) model for the progress-variable SGS variance field is trained on a canonical premixed turbulent flame and evaluated a priori on a significantly more complex slot burner jet flame. Despite the extensive differences between the two configurations, the CNN generalizes well and outperforms existing algebraic models. Conditions for this successful generalization are discussed, including the effect of the filter size and flame–turbulence interaction parameters. The CNN is then integrated into an analytical reaction rate closure relying on a single-step chemical source term formulation and a presumed beta PDF (probability density function) approach. The proposed closure is able to accurately recover filtered reaction rate values on both training and generalization flames. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

31 pages, 141229 KiB

Open AccessArticle

Experimental and Numerical Analysis on Two-Phase Induced Low-Speed Pre-Ignition

by Norbert Zöbinger, Thorsten Schweizer, Thomas Lauer, Heiko Kubach and Thomas Koch

Energies 2021, 14(16), 5063; https://doi.org/10.3390/en14165063 - 17 Aug 2021

Cited by 3 | Viewed by 1959

Abstract

The root cause of the initial low-speed pre-ignition (LSPI) is not yet clarified. The literature data suggest that a two-phase phenomenon is most likely triggering the unpredictable premature ignitions in highly boosted spark-ignition engines. However, there are different hypotheses regarding the actual initiator, [...] Read more.

The root cause of the initial low-speed pre-ignition (LSPI) is not yet clarified. The literature data suggest that a two-phase phenomenon is most likely triggering the unpredictable premature ignitions in highly boosted spark-ignition engines. However, there are different hypotheses regarding the actual initiator, whether it is a detached liquid oil-fuel droplet or a solid-like particle from deposits. Therefore, the present work investigates the possibility of oil droplet-induced pre-ignitions using a modern downsized engine with minimally invasive endoscopic optical accessibility incorporating in-cylinder lubrication oil detection via light-induced fluorescence. This setup enables the differentiation between liquid and solid particles. Furthermore, the potential of hot solid particles to initiate an ignition under engine-relevant conditions is analyzed numerically. To do so, the particle is generalized as a hot surface transferring heat to the reactive ambient gas phase. The gas-phase reactivity is represented as a TRF/air mixture based on RON/MON specifications of the investigated fuel. The chemical processes are predicted using a semi-detailed reaction mechanism, including 137 species and 633 reactions in a 2D CFD simulation framework. In the optical experiments, no evidence of a liquid oil droplet-induced pre-ignition could be found. Nevertheless, all observed pre-ignitions had a history of flying light-emitting objects. There are strong hints towards solid-like deposit LSPI initiation. The application of the numerical methodology to mean in-cylinder conditions of an LSPI prone engine operation point reveals that particles below 1000 K are not able to initiate a pre-ignition. A sensitivity analysis of the thermodynamic boundary conditions showed that the particle temperature is the most decisive parameter on the calculated ignition delay time. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

25 pages, 6542 KiB

Open AccessArticle

Numerical Investigation of the Turbulent Flame Propagation in Dual Fuel Engines by Means of Large Eddy Simulation

by Jens Frühhaber and Thomas Lauer

Energies 2021, 14(16), 5036; https://doi.org/10.3390/en14165036 - 17 Aug 2021

Cited by 1 | Viewed by 1884

Abstract

Dual fuel combustion depicts a possible alternative to reduce emissions from large engines and is characterized by injecting a small amount of diesel fuel into a lean natural gas–air mixture. Thereby, the presence of autoignition, diffusive and premixed combustion determine the high complexity [...] Read more.

Dual fuel combustion depicts a possible alternative to reduce emissions from large engines and is characterized by injecting a small amount of diesel fuel into a lean natural gas–air mixture. Thereby, the presence of autoignition, diffusive and premixed combustion determine the high complexity of this process. In this work, an Extended Coherent Flame Model was adapted to consider the effect of natural gas on the ignition delay time. This model was afterward utilized to simulate 25 consecutive engine cycles employing LES. In this framework, the ensemble-average flow field was compared to a RANS solution to assess the advantages of LES in terms of the prediction of the in-cylinder flow field. A detailed investigation of the heat release characteristic showed that natural gas already highly contributes to the heat release at the beginning of combustion. Furthermore, a methodology to investigate the turbulent combustion regimes was utilized. It could be ascertained that the combustion mainly occurs in the regime of thin reaction zones. Possible triggers of cycle-to-cycle variations were determined in the velocity fluctuations in the cylinder axis direction and the flame formation in the gaps between the spray plume. The findings support the understanding of dual fuel combustion and serve as a basis for developing future combustion models. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

17 pages, 4991 KiB

Open AccessArticle

A Data-Driven Methodology for the Simulation of Turbulent Flame Speed across Engine-Relevant Combustion Regimes

by Alessandro d’Adamo, Clara Iacovano and Stefano Fontanesi

Energies 2021, 14(14), 4210; https://doi.org/10.3390/en14144210 - 12 Jul 2021

Cited by 1 | Viewed by 2422

Abstract

Turbulent combustion modelling in internal combustion engines (ICEs) is a challenging task. It is commonly synthetized by incorporating the interaction between chemical reactions and turbulent eddies into a unique term, namely turbulent flame speed

s_{T}

. The task is very complex considering [...] Read more.

Turbulent combustion modelling in internal combustion engines (ICEs) is a challenging task. It is commonly synthetized by incorporating the interaction between chemical reactions and turbulent eddies into a unique term, namely turbulent flame speed

s_{T}

. The task is very complex considering the variety of turbulent and chemical scales resulting from engine load/speed variations. In this scenario, advanced turbulent combustion models are asked to predict accurate burn rates under a wide range of turbulence–flame interaction regimes. The framework is further complicated by the difficulty in unambiguously evaluating in-cylinder turbulence and by the poor coherence of turbulent flame speed (

s_{T}

) measurements in the literature. Finally, the simulated

s_{T}

from combustion models is found to be rarely assessed in a rigorous manner. A methodology is presented to objectively measure the simulated

s_{T}

by a generic combustion model over a range of engine-relevant combustion regimes, from

D a

= 0.5 to

D a

= 75 (i.e., from the thin reaction regime to wrinkled flamelets). A test case is proposed to assess steady-state burn rates under specified turbulence in a RANS modelling framework. The methodology is applied to a widely adopted combustion model (ECFM-3Z) and the comparison of the simulated

s_{T}

with experimental datasets allows to identify modelling improvement areas. Dynamic functions are proposed based on turbulence intensity and Damköhler number. Finally, simulations using the improved flame speed are carried out and a satisfactory agreement of the simulation results with the experimental/theoretical correlations is found. This confirms the effectiveness and the general applicability of the methodology to any model. The use of grid/time resolution typical of ICE combustion simulations strengthens the relevance of the proposed dynamic functions. The presented analysis allows to improve the adherence of the simulated burn rate to that of literature turbulent flames, and it unfolds the innovative possibility to objectively test combustion models under any prescribed turbulence/flame interaction regime. The solid data-driven representation of turbulent combustion physics is expected to reduce the tuning effort in ICE combustion simulations, providing modelling robustness in a very critical area for virtual design of innovative combustion systems. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

28 pages, 43260 KiB

Open AccessArticle

A Neural Network-Inspired Matrix Formulation of Chemical Kinetics for Acceleration on GPUs

by Shivam Barwey and Venkat Raman

Energies 2021, 14(9), 2710; https://doi.org/10.3390/en14092710 - 9 May 2021

Cited by 22 | Viewed by 3236

Abstract

High-fidelity simulations of turbulent flames are computationally expensive when using detailed chemical kinetics. For practical fuels and flow configurations, chemical kinetics can account for the vast majority of the computational time due to the highly non-linear nature of multi-step chemistry mechanisms and the [...] Read more.

High-fidelity simulations of turbulent flames are computationally expensive when using detailed chemical kinetics. For practical fuels and flow configurations, chemical kinetics can account for the vast majority of the computational time due to the highly non-linear nature of multi-step chemistry mechanisms and the inherent stiffness of combustion chemistry. While reducing this cost has been a key focus area in combustion modeling, the recent growth in graphics processing units (GPUs) that offer very fast arithmetic processing, combined with the development of highly optimized libraries for artificial neural networks used in machine learning, provides a unique pathway for acceleration. The goal of this paper is to recast Arrhenius kinetics as a neural network using matrix-based formulations. Unlike ANNs that rely on data, this formulation does not require training and exactly represents the chemistry mechanism. More specifically, connections between the exact matrix equations for kinetics and traditional artificial neural network layers are used to enable the usage of GPU-optimized linear algebra libraries without the need for modeling. Regarding GPU performance, speedup and saturation behaviors are assessed for several chemical mechanisms of varying complexity. The performance analysis is based on trends for absolute compute times and throughput for the various arithmetic operations encountered during the source term computation. The goals are ultimately to provide insights into how the source term calculations scale with the reaction mechanism complexity, which types of reactions benefit the GPU formulations most, and how to exploit the matrix-based formulations to provide optimal speedup for large mechanisms by using sparsity properties. Overall, the GPU performance for the species source term evaluations reveals many informative trends with regards to the effect of cell number on device saturation and speedup. Most importantly, it is shown that the matrix-based method enables highly efficient GPU performance across the board, achieving near-peak performance in saturated regimes. Full article

(This article belongs to the Special Issue Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Computational and Data-Driven Modeling of Turbulent Combustion and Engine Combustion Dynamics

Share This Special Issue

Special Issue Editor

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Published Papers (8 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI