Combustion machine learning: Principles, progress and prospects

Ihme, Matthias; Chung, Wai Tong; Mishra, Aashwin

doi:10.1016/j.pecs.2022.101010

Cited by 118 publications

(49 citation statements)

References 616 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We following common practice and split each dataset into training, validation, and testing sets with a ratio of 80:10:10 [18], resulting in a total of 8000 unique training fire sequences, 1,000 validation fire sequences and 1,000 testing fire sequences. Each model was trained only on data from a single dataset.…”

Section: Additional Training Detailsmentioning

confidence: 99%

“…In the past decade, the potential of machine learning (ML) methods for application to wildfires has been recognized, with ML techniques used across tasks such as fuel characterization, risk assessment, fire behavior modeling, and fire management [17,18]. While there are various approaches within the ML community, the sub-field of deep learning with the development of deep neural network (DNN) has led to breakthroughs in many domains [19].…”

mentioning

confidence: 99%

See 1 more Smart Citation

Recurrent Convolutional Deep Neural Networks for Modeling Time-Resolved Wildfire Spread Behavior

Burge¹,

Bonanni²,

Lily³

et al. 2022

Preprint

View full text Add to dashboard Cite

The increasing incidence and severity of wildfires underscores the necessity of accurately predicting their behavior. While high-fidelity models derived from first principles offer physical accuracy, they are too computationally expensive for use in real-time fire response. Lowfidelity models sacrifice some physical accuracy and generalizability via the integration of empirical measurements, but enable real-time simulations for operational use in fire response.Machine learning techniques offer the ability to bridge these objectives by learning firstprinciples physics while achieving computational speedup. While deep learning approaches have demonstrated the ability to predict wildfire propagation over large time periods, timeresolved fire-spread predictions are needed for active fire management. In this work, we evaluate the ability of deep learning approaches in accurately modeling the time-resolved dynamics of wildfires. We use an autoregressive process in which a convolutional recurrent deep learning model makes predictions that propagate a wildfire over 15 minute increments.We demonstrate the model in application to three simulated datasets of increasing complexity, containing both field fires with homogeneous fuel distribution as well as real-world topologies sampled from the California region of the United States. We show that even after 100 autoregressive predictions representing more than 24 hours of simulated fire spread, the resulting models generate stable and realistic propagation dynamics, achieving a Jaccard score between 0.89 and 0.94 when predicting the resulting fire scar.

show abstract

Section: Additional Training Detailsmentioning

confidence: 99%

mentioning

confidence: 99%

Recurrent Convolutional Deep Neural Networks for Modeling Time-Resolved Wildfire Spread Behavior

Burge¹,

Bonanni²,

Lily³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In the past decades, considerable attention has been paid to LES closure model developments through algebraic-based equations [2]. Recently, Deep Learning (DL) techniques have performed well in modeling non-linear flow interactions, and thus hold the promise of advancing modeling and analyzing the intricate structures associated with turbulent reacting flows [3,4]. The usage of DL methods for LES closure models has been identified as one of its key applications by the fluid dynamics community [5].…”

Section: Introductionmentioning

confidence: 99%

“…In this view, Ledig et al [12] pro-posed a novel architecture, the Super-Resolution Generative Adversarial Network (SRGAN), using perceptual and adversarial losses to favor outputs residing on the manifold of natural images. Thanks to their distinguished capabilities, several of those architectures have been recently applied for closure modeling [4]. In the context of LES modeling, the work of Fukami et al [13] demonstrated the ability to use deep CNNs to super-resolve three-dimensional incompressible turbulent flows solely from the coarsegrained data fields, enhancing subfilter physical structures.…”

Section: Introductionmentioning

confidence: 99%

“…Another potential drawback of data-based ML models is that highly-resolved data needed for the training is available only at a relatively low Re number compared to real applications (e.g., gas-turbine combustors) due to the prohibitive cost of DNS. Without generalization capabilities, the supervised data-driven models are limited to physical conditions for which DNS data or highly-resolved LES is available [4].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Investigation of the generalization capability of a generative adversarial network for large eddy simulation of turbulent premixed reacting flows

Nista

Schumann

Grenga

et al. 2023

Proceedings of the Combustion Institute

View full text Add to dashboard Cite

On the Use of Machine Learning for Subgrid Scale Filtered Density Function Modelling in Large Eddy Simulations of Combustion Systems

Iavarone

Yang

et al. 2023

Lecture Notes in Energy

View full text Add to dashboard Cite

The application of machine learning algorithms to model subgrid-scale filtered density functions (FDFs), required to estimate filtered reaction rates for Large Eddy Simulation (LES) of chemically reacting flows, is discussed in this chapter. Three test cases, i.e., a low-swirl premixed methane-air flame, a MILD combustion of methane-air mixtures, and a kerosene spray turbulent flame, are presented. The scalar statistics in these test cases may not be easily represented using the commonly used presumed shapes for modeling FDFs of mixture fraction and progress variable. Hence, the use of ML methods is explored. Particularly, deep neural network (DNN) to infer joint FDFs of mixture fraction and progress variable is reviewed here. The Direct Numerical Simulation (DNS) datasets employed to train the DNNs in each test case are described. The DNN performances are shown and compared to typical presumed probability density function (PDF) models. Finally, this chapter examines the advantages and caveats of the DNN-based approach.

show abstract

Combustion machine learning: Principles, progress and prospects

Cited by 118 publications

References 616 publications

Recurrent Convolutional Deep Neural Networks for Modeling Time-Resolved Wildfire Spread Behavior

Recurrent Convolutional Deep Neural Networks for Modeling Time-Resolved Wildfire Spread Behavior

Investigation of the generalization capability of a generative adversarial network for large eddy simulation of turbulent premixed reacting flows

On the Use of Machine Learning for Subgrid Scale Filtered Density Function Modelling in Large Eddy Simulations of Combustion Systems

Contact Info

Product

Resources

About