A deep learning framework for hydrogen-fueled turbulent combustion simulation

An, Jian; Wang, Hanyi; Liu, Bing; Luo, Kai H.; Qin, Fei; He, Guo Qiang

doi:10.1016/j.ijhydene.2020.04.286

Cited by 29 publications

(6 citation statements)

References 39 publications

(40 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many researchers are still working on speeding up CFD techniques for detailed mechanisms. Despite the development of new combustion models or mechanism reduction, some possible approaches may be the utilization of the field programmable gate array (FPGA), the graphics processing unit (GPU)-accelerated chemistry solver, − and the application of machine learning. − It should be noted that the GPU-based solver and machine learning are both software-based approaches, while the FPGA approach adjusts the hardware to solve a specific CFD problem instead of software optimization. For instance, in Ebrahimi and Zandsalimy’s work, the authors found that the FPGA improved the solution speed up to 20 times faster in the case of the Laplace equation (compared to the conventional CPU).…”

Section: Coupling Kinetic Model With Cfdmentioning

confidence: 99%

A Review on Detailed Kinetic Modeling and Computational Fluid Dynamics of Thermochemical Processes of Solid Fuels

et al. 2021

View full text Add to dashboard Cite

This review presents the recent advances in computational approaches to understand thermochemical reactions of solid fuels with minimized empirical factors. It includes studies on kinetic modeling of gas-phase reactions of multicomponent molecular mixtures of volatiles derived from primary pyrolysis of solid fuels using a database for elementary reactions. Mechanistic studies to model the devolatilization of biomass and coal upon heating are also mentioned. The efforts to integrate the kinetic model with computational fluid dynamics to understand the flow and heat transfer behavior of industrial reactors for solid fuel conversions, including gasification and combustion, are reviewed. These advances are primarily based on the conventional forward analysis that provides a deeper understanding of chemically reacting flows of fuels undergoing thermochemical reactions. For the further development of the thermochemical conversion technology, it can be expected to develop alternative approaches such as inverse analysis to determine the optimal reactor design and operating conditions.

show abstract

Section: Coupling Kinetic Model With Cfdmentioning

confidence: 99%

A Review on Detailed Kinetic Modeling and Computational Fluid Dynamics of Thermochemical Processes of Solid Fuels

et al. 2021

View full text Add to dashboard Cite

show abstract

“…While traditional RNNs can preserve temporal relationships, they overlook the acquisition of spatial features, which are more critical for this type of task. According to relevant literature, when dealing with similar spatiotemporal data, spatial features take precedence over temporal features, and most DL models used for reconstructing flow fields rely on CNN. ,, In contrast, RNNs are more suitable for handling numerical data. A feasible approach is to decompose the image into a combination of a large number of pixels and treat them as numerical data, but this method requires an unacceptable amount of memory capacity.…”

Section: Introductionmentioning

confidence: 99%

Predicting Spatiotemporal Distributions in a Bubbling Fluidized Bed for Biomass Fast Pyrolysis Using Convolutional Neural Networks

Zhong,

Yu,

Zhang

et al. 2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Bubbling fluidized-bed biomass fast pyrolysis is a crucial technology for carbon neutrality and sustainability, and computational fluid dynamics (CFD) is one of the promising approaches to investigate and optimize bubbling fluidized-bed biomass fast pyrolysis. However, traditional CFD is still computationally costly for bubbling fluidized-bed biomass fast pyrolysis, especially for spatiotemporal transport-reaction behaviors, which are critical to clarifying intrinsic characteristics and optimizing operations. To address this issue, a deep learning (DL) model centered on convolutional neural networks was developed based on CFD results to efficiently predict spatiotemporal distributions of quantities of each phase in a bubbling fluidized bed for biomass fast pyrolysis. Input of the DL model is a sequence of spatiotemporal distributions, and only an initial input is required to generate continuous outputs. The model was optimized by adjusting four typical parameters, i.e., length of input sequence, number of neurons, learning rate, and prediction step size. Accuracy of short-term prediction (10 frames) and stability of long-term prediction (1000 frames) were analyzed as well as the relationship between time-averaged distributions and prediction length. It was found that with satisfactory accuracy, several orders of magnitude increase in computation efficiency can be realized. Thus, the developed model paves the way for low-cost and high-accuracy simulations of biomass fast pyrolysis.

show abstract

“…The field of combustion has already shown promising research results through the application of artificial intelligence. For example, An et al [ 33 ] leveraged a neural network as an auxiliary solver for computational fluid dynamics methods to predict the distribution of flow and temperature fields in a combustor. Zhou et al [ 34 ] investigated the use of deep learning based on time-averaged flame images to monitor combustion instabilities.…”

Section: Introductionmentioning

confidence: 99%

Predicting NOx Distribution in a Micro Rich–Quench–Lean Combustor Using a Variational Autoencoder

Yan

Fan

Zhang

2023

Entropy

View full text Add to dashboard Cite

Micro gas turbines are widely used in distributed power generation systems. However, the combustion of gas turbine combustors produces a large amount of nitrogen oxides (NOx), which pollute the environment and endanger human life. To reduce environmental pollution, low-emission combustors have been developed. In recent years, there has been an increasing focus on the use of low-heat-value gas fuels, and it is necessary to study the NOx emissions from low heat value gas fuel combustors. Data-driven deep learning methods have been used in many fields in recent years. In this study, a variational autoencoder was introduced for the prediction of NOx production inside the combustor. The combustor used was a micro rich–quench–lean combustor designed by the research group using coal bed gas as a fuel. The internal NO distribution contour was obtained as the dataset using simulation methods, with a size of 60 images. The model architecture parameters were obtained through hyperparameter exploration using the grid search method. The model accurately predicted the distribution of NO inside the combustor. The method can be applied in the prediction of a wider range of parameters and offers a new way of designing combustors for the power industry.

show abstract

A deep learning framework for hydrogen-fueled turbulent combustion simulation

Cited by 29 publications

References 39 publications

A Review on Detailed Kinetic Modeling and Computational Fluid Dynamics of Thermochemical Processes of Solid Fuels

A Review on Detailed Kinetic Modeling and Computational Fluid Dynamics of Thermochemical Processes of Solid Fuels

Predicting Spatiotemporal Distributions in a Bubbling Fluidized Bed for Biomass Fast Pyrolysis Using Convolutional Neural Networks

Predicting NOx Distribution in a Micro Rich–Quench–Lean Combustor Using a Variational Autoencoder

Contact Info

Product

Resources

About