相关参考文献
注意：仅列出部分参考文献，下载原文获取全部文献信息。
Article
Thermodynamics
Kamila Zdybal et al.
Summary: Reduced-order models (ROMs) for turbulent combustion aim to describe complex reacting flows with a small number of effective parameters. This study proposes a quantitative manifold-informed method for selecting a subset of state variables to improve the quality of low-dimensional data representations. The authors demonstrate that a mixture of major and minor species can be beneficial in reducing non-uniqueness and spatial gradients in the dependent variable space.
PROCEEDINGS OF THE COMBUSTION INSTITUTE
(2023)
Article
Physics, Fluids & Plasmas
Clement Scherding et al.
Summary: In this paper, a model-agnostic machine-learning technique is proposed to extract a reduced thermochemical model of a gas mixture from a library, reducing the cost of simulating hypersonic flows. By clustering and generating surrogate surfaces, different thermochemical states are handled, and the method is validated to improve solver performance by up to 70% while maintaining accuracy.
PHYSICAL REVIEW FLUIDS
(2023)
Article
Engineering, Aerospace
Ali C. Ispir et al.
Summary: Dual-mode ramjet/scramjet engines are commonly preferred air-breathing propulsion systems for hypersonic aircraft. Modeling the fuel-air mixing process is a main challenge in their design to optimize engine performance. Machine learning models, such as artificial neural networks, can be used for multi-objective optimization of design variables. Experimental data can be used to build regression models that predict mixing conditions in reduced-order modeling studies for estimating engine performance.
Article
Thermodynamics
Elizabeth Armstrong et al.
Summary: Tabulated chemistry models are widely used in simulating large-scale turbulent fires. Artificial neural networks often fail to converge and provide accurate results, whereas partition of unity networks offer higher accuracy and lower memory usage.
COMBUSTION SCIENCE AND TECHNOLOGY
(2022)
Article
Thermodynamics
Bruce A. Perry et al.
Summary: Many modeling approaches in LES of turbulent combustion use low-dimensional manifold projection to reduce computational cost. This study proposes a new approach that extends existing methods by modifying the neural network to simultaneously encode manifold variables, nonlinear mapping, and subfilter closure. The approach shows improved performance compared to PCA-based models.
COMBUSTION AND FLAME
(2022)
Article
Multidisciplinary Sciences
Kamila Zdybal et al.
Summary: Reduced-order modeling is a method to describe complex systems with high state-space dimensionality using a small number of parameters. This paper presents a quantitative metric for characterizing the quality of manifold topologies. The metric considers non-uniqueness and spatial gradients in physical quantities of interest, and can be used in optimization algorithms to find optimized low-dimensional projections.
SCIENTIFIC REPORTS
(2022)
Article
Thermodynamics
Elizabeth Armstrong et al.
Summary: Effective dimension reduction is crucial for large-scale simulation of high-dimensional dynamical systems, and accurate reconstruction of nonlinear functions is key. Existing manifold quality assessments have limitations in considering features such as sharp gradients or non-uniqueness.
COMBUSTION THEORY AND MODELLING
(2021)
Article
Computer Science, Software Engineering
Kamila Zdybal et al.
