BLASTNet: A call for community-involved big data in combustion machine learning

Chung, Wai Tong; Jung, Ki Sung; Chen, Jacqueline H.; Ihme, Matthias

doi:10.1016/j.jaecs.2022.100087

Cited by 5 publications

(4 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This culture could be emulated for fluid mechanics studies, which would result in numerous benchmark studies for the fluid mechanics community. However, significant strides in (i) forming community-accepted datasets [102] and (ii) organizing a community-wide benchmarking study will need to be accomplished before such a challenge can be met. Numerous efforts discussed in Section 2.4 present a step toward addressing these challenges.…”

Section: Benchmarking Existing Piml Methodsmentioning

confidence: 99%

“…To deal with the large cost and size of DNS datasets, crowd-sourcing and public repositories such as Kaggle [101] could be employed to improve access to high-fidelity ML training data. For example, BLASTNet [102], a centralized web platform for indexing DNS data with consistent data formats, has recently been developed. In this approach, members from the research community contribute to individual DNS datasets that are stored in Kaggle repositories (each with O(100) GB storage limit), with download links and metadata consolidated in a single webpage for easy accessibility.…”

Section: Open-source Piml Resourcesmentioning

confidence: 99%

See 1 more Smart Citation

A Review of Physics-Informed Machine Learning in Fluid Mechanics

et al. 2023

View full text Add to dashboard Cite

Physics-informed machine-learning (PIML) enables the integration of domain knowledge with machine learning (ML) algorithms, which results in higher data efficiency and more stable predictions. This provides opportunities for augmenting—and even replacing—high-fidelity numerical simulations of complex turbulent flows, which are often expensive due to the requirement of high temporal and spatial resolution. In this review, we (i) provide an introduction and historical perspective of ML methods, in particular neural networks (NN), (ii) examine existing PIML applications to fluid mechanics problems, especially in complex high Reynolds number flows, (iii) demonstrate the utility of PIML techniques through a case study, and (iv) discuss the challenges and opportunities of developing PIML for fluid mechanics.

show abstract

Section: Benchmarking Existing Piml Methodsmentioning

confidence: 99%

Section: Open-source Piml Resourcesmentioning

confidence: 99%

A Review of Physics-Informed Machine Learning in Fluid Mechanics

et al. 2023

View full text Add to dashboard Cite

show abstract

“…For the broader use of the transcritical DNS database by the turbulence and combustion community, we have made this database open-access on https://blastnet. github.io/index.html (Chung et al 2022). The flow has a reduced bulk pressure (defined as p r = p/p c with p c the critical pressure) of 1.14, with the temperatures for the hot wall and the cold wall kept spatially and temporally constant.…”

Section: Problem Formulationmentioning

confidence: 99%

Small-scale turbulent characteristics in transcritical wall-bounded flows

Li,

Zhang,

Bai

et al. 2024

J. Fluid Mech.

View full text Add to dashboard Cite

Considerable efforts have been devoted to the understanding of the small-scale characteristics in turbulent flows. While the universality of small-scale quantities has been established for incompressible flows, their extension to high-pressure transcritical flows remains an open area of research. To address this question, we investigate the real-fluid thermodynamic effects on small-scale velocity statistics of high-pressure transcritical wall-bounded turbulence. We show that in the locally isotropic region for transcritical flows, low-order moments of small-scale statistics collapse for all cases and Kolmogorov's (1941) theory holds. However, real-fluid thermodynamic effects introduce deviations in the tails of the probability density function for the velocity derivative and, consequently, high-order moments of velocity gradients and dissipation rate in transcritical flows cannot collapse in the locally isotropic region. Analysis of the intermittency shows that the low-order structure functions in transcritical flows follow extended self-similarity, while the dependence of the intermittency factor on real-fluid effects is observed for high-order structure functions. The real-fluid effects on intermittency are explained by turbulent structures related to rare events. Additionally, the dissipation rate moments for transcritical flows follow a universal scaling with Reynolds number, and the scaling exponents are different from those of incompressible flows. These results extend the small-scale universality in incompressible flows (Schumacher et al., Proc. Natl Acad. Sci. USA, vol. 111, 2014, pp. 10961–10965) to realistic flows with significant changes in thermodynamic properties, and provide a physical underpinning of the scaling laws of small-scale statistics at transcritical conditions.

show abstract

“…Some recent attempts have been made to create databases and benchmark cases (e.g., the BLASTNet 15 ). A recent review showed that more than 200 DNS studies on turbulent reacting flows over the last two decades were published during the last two decades.…”

Section: The Role Of Data In the Development Of Sustainable Combustio...mentioning

confidence: 99%