A multi-scale sampling method for accurate and robust deep neural network to predict combustion chemical kinetics
AI for Science Institute（北京科学智能研究院）；Princeton University（普林斯顿大学）；Shanghai Jiao Tong University（上海交通大学）；Peking University（北京大学）；Shanghai Center for Brain Science and Brain-Inspired Technology（上海脑科学与类脑研究中心）
Tianhan Zhang, Yuxiao Yi, Yifan Xu, Zhi X. Chen, Yaoyu Zhang, Weinan E, Zhi-Qin John Xu
Machine learning has long been considered as a black box for predicting combustion chemical kinetics due to the extremely large number of parameters and the lack of evaluation standards and reproducibility. The current work aims to understand two basic questions regarding the deep neural network (DNN) method: what data the DNN needs and how general the DNN method can be. Sampling and preprocessing determine the DNN training dataset, further affect DNN prediction ability. The current work proposes using Box-Cox transformation (BCT) to preprocess the combustion data. In addition, this work compares different sampling methods with or without preprocessing, including the Monte Carlo method, manifold sampling, generative neural network method (cycle-GAN), and newly-proposed multi-scale sampling. Our results reveal that the DNN trained by the manifold data can capture the chemical kinetics in limited configurations but cannot remain robust toward perturbation, which is inevitable for the DNN coupled with the flow field. The Monte Carlo and cycle-GAN samplings can cover a wider phase space but fail to capture small-scale intermediate species, producing poor prediction results. A three-hidden-layer DNN, based on the multi-scale method without specific flame simulation data, allows predicting chemical kinetics in various scenarios and being stable during the temporal evolutions. This single DNN is readily implemented with several CFD codes and validated in various combustors, including (1). zero-dimensional autoignition, (2). one-dimensional freely propagating flame, (3). two-dimensional jet flame with triple-flame structure, and (4). three-dimensional turbulent lifted flames. The results demonstrate the satisfying accuracy and generalization ability of the pre-trained DNN. The Fortran and Python versions of DNN and example code are attached in the supplementary for reproducibility.
Direct fired supercritical CO2 (sCO2) power cycles allow for the combustion of gaseous fuels under oxyfuel conditions with inherent carbon capture. As the CO2 is captured intrinsically, the efficiency penalty of capture on the overall plant is small, meaning that power plants achieve a similar efficiency to traditional fossil fuel power plants without carbon capture and storage. However, at high pressures and in large dilutions of CO2, combustion mechanisms are poorly understood. Therefore, in this paper sensitivity and quantitative analysis of four established chemical kinetic mechanisms have…
Combustion kinetic modeling is an integral part of combustion simulation, and extensive studies have been devoted to developing both high fidelity and computationally affordable models. Despite these efforts, modeling combustion kinetics is still challenging due to the demand for expert knowledge and optimization against experiments, as well as the lack of understanding of the associated uncertainties. Therefore, data-driven approaches that enable efficient discovery and calibration of kinetic models have received much attention in recent years, the core of which is the optimization based on b…
East China Normal University（华东师范大学）；New York University（纽约大学）
Understanding and prediction of the chemical reactions are fundamental demanding in the study of many complex chemical systems. Reactive molecular dynamics (MD) simulation has been widely used for this purpose as it can offer atomic details and can help us better interpret chemical reaction mechanisms. In this study, two reference datasets were constructed and corresponding neural network (NN) potentials were trained based on them. For given large-scale reaction systems, the NN potentials can predict the potential energy and atomic forces of DFT precision, while it is orders of magnitude faste…
Accurate low dimension chemical kinetic models for methane are an essential component in the design of efficient gas turbine combustors. Kinetic models coupled to computational fluid dynamics (CFD) provide quick and efficient ways to test the effect of operating conditions, fuel composition and combustor design compared to physical experiments. However, detailed chemical kinetic models are too computationally expensive for use in CFD. We propose a novel data orientated three-step methodology to produce compact models that replicate a target set of detailed model properties to a high fidelity. …
Clarifying NH2 + O(3P) Reaction Dynamics: A Full-Dimensional MRCI, Machine-Learned PES Unravels High-Temperature Kinetics
MIIT Key Laboratory of Semiconductor Microstructure and Quantum Sensing, Department of Applied Physics, School of Physics, Nanjing University of Science and Technology（南京理工大学应用物理学院）
Combustion modes in locally stratified dual-fuel (DF) mixtures are numerically investigated for methanol n-dodecane blends under engine-relevant pressures. In the studied constant-volume numerical setup, methanol acts as a background low-reactivity fuel (LRF) while n-dodecane serves as high-reactivity fuel (HRF), controlling local ignition delay time. The spatial distribution of n-dodecane is modeled as a sinusoidal function parametrized by stratification amplitude (Y) and wavelength (0.01 mm<$λ$<15 mm). In contrast, methanol is assumed to be fully premixed with air at equivalence ratio 0.8. F…
Shock tube experiments, paired with precision laser diagnostics, are ideal venues to provide kinetics data critically needed for the development, validation and optimization of modern combustion kinetics models. However, to design sensitive, accurate, feasible and information-rich experiments that may yield such data often requires sophisticated planning. This study presents a mathematical framework and quantitative approach to guide such experimental design, namely a method to pin-point the optimal conditions for specific experimentation under realistic constraints of the shock tubes and diag…
Bulk nanobubbles (NBs) generated electrochemically by short voltage pulses of alternating polarity behave differently from those produced by regular methods. Only bubbles smaller than $200\;$nm are formed in the process and their concentration is very high. Moreover, the bubbles containing both H2 and O2 gases disappear quickly via the combustion reaction, although the reaction in such a small volume cannot happen according to the classical combustion theory. Experimental facts about these unusual NBs are reviewed and current understanding of the observed phenomena is provided. Visualisation m…
Chemical kinetics mechanisms are essential for understanding, analyzing, and simulating complex combustion phenomena. In this study, a Neural Ordinary Differential Equation (Neural ODE) framework is employed to optimize kinetics parameters of reaction mechanisms. Given experimental or high-cost simulated observations as training data, the proposed algorithm can optimally recover the hidden characteristics in the data. Different datasets of various sizes, types, and noise levels are tested. A classic toy problem of stiff Robertson ODE is first used to demonstrate the learning capability, effici…
A wide variety of real life complex networks are prohibitively large for modeling, analysis and control. Understanding the structure and dynamics of such networks entails creating a smaller representative network that preserves its relevant topological and dynamical properties. While modern machine learning methods have enabled identification of governing laws for complex dynamical systems, their inability to produce white-box models with sufficient physical interpretation renders such methods undesirable to domain experts. In this paper, we introduce a hybrid black-box, white-box approach for…
Recently developed physics-informed neural network (PINN) has achieved success in many science and engineering disciplines by encoding physics laws into the loss functions of the neural network, such that the network not only conforms to the measurements, initial and boundary conditions but also satisfies the governing equations. This work first investigates the performance of PINN in solving stiff chemical kinetic problems with governing equations of stiff ordinary differential equations (ODEs). The results elucidate the challenges of utilizing PINN in stiff ODE systems. Consequently, we empl…
Massachusetts Institute of Technology（麻省理工学院）；University of California, Berkeley（加州大学伯克利分校）
Modeling the burning processes of biomass such as wood, grass, and crops is crucial for the modeling and prediction of wildland and urban fire behavior. Despite its importance, the burning of solid fuels remains poorly understood, which can be partly attributed to the unknown chemical kinetics of most solid fuels. Most available kinetic models were built upon expert knowledge, which requires chemical insights and years of experience. This work presents a framework for autonomously discovering biomass pyrolysis kinetic models from thermogravimetric analyzer (TGA) experimental data using the rec…
Inclusion of sulfur chemistry in a validated C/H/O/N chemical network: identification of key C/S coupling pathways
Université Paris Cité and Univ Paris Est Creteil, CNRS, LISA（巴黎西岱大学与巴黎东克雷泰伊大学，法国国家科学研究中心，LISA）；Astrophysics Group, University of Exeter（埃克塞特大学天体物理学组）；Université de Lorraine, CNRS, LRGP（洛林大学，法国国家科学研究中心，LRGP）；Department of Physics and Astronomy, University College London（伦敦大学学院物理与天文系）
Chemical kinetics plays an important role in governing the thermal evolution in reactive flows problems. The possible interactions between chemical species increase drastically with the number of species considered in the system. Various ways have been proposed before to simplify chemical networks with an aim to reduce the computational complexity of the chemical network. These techniques oftentimes require domain-knowledge experts to handcraftedly identify important reaction pathways and possible simplifications. Here, we propose a combination of autoencoder and neural ordinary differential e…
A New Data-Driven Sparse-Learning Approach to Study Chemical Reaction Networks
Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI 48109--2125, USA（电气工程与计算机科学系，密歇根大学，安娜堡，MI 48109--2125，美国）；Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI 48109--2125, USA（机械工程系，密歇根大学，安娜堡，MI 48109--2125，美国）；Departments of Chemical Engineering, Biomedical Engineering, Macromolecular Science and Engineering, Biophysics Program, University of Michigan, Ann Arbor, MI 48109--2125, USA（化学工程、生物医学工程、大分子科学与工程、生物物理项目系，密歇根大学，安娜堡，MI 48109--2125，美国）
In many scientific fields, there is an interest in understanding the way in which complex chemical networks evolve. The chemical networks which researchers focus upon, have become increasingly complex and this has motivated the development of automated methods for exploring chemical reactivity or conformational change in a black box manner, harnessing modern computing resources to automate mechanism discovery. In this work we present a new approach to automated mechanism generation implemented which couples molecular dynamics and statistical rate theory to automatically find kinetically import…
The theory of chemical kinetics form the basis to describe the dynamics of chemical systems. Owing to physical and thermodynamic constraints, chemical reaction systems possess various structures, which can be utilized to characterize important physical properties of the systems. In this work, we reveal the Hessian geometry which underlies chemical reaction systems and demonstrate how it originates from the interplay of stoichiometric and thermodynamic constraints. Our derivation is based on kinetics, we assume the law of mass action and characterize the equilibrium states by the detailed balan…
The classical mass action law in chemical kinetics is put into the context of multiscale thermodynamics.Despite the purely dissipative character of the classical mass action law, inertial effects also play a role in chemical kinetics. Therefore, the kinetics is extended to an enlarged state space with reaction rates as new state variables and bringing inertial effects. The dynamics is then lifted to the Liouville description within kinetic theory on the enlarged state space in order to include fluctuations. Subsequently, the kinetic theory is reduced to its first and second moments, leading to…
Recent experiments have suggested that ground state chemical kinetics can be suppressed or enhanced by coupling the vibrational degrees of freedom of a molecular system with a radiation mode inside an optical cavity. Experiments show that the chemical rate is strongly modified when the photon frequency is close to characteristic vibrational frequencies. The origin of this remarkable effect remains unknown. In this work, we develop an analytical rate theory for cavity-modified ground state chemical kinetics based on the Pollak-Grabert-Hänggi rate theory. Unlike previous work, our theory covers …
Using the formalism of the classical nucleation theory, we derive a novel kinetic equation for the size and composition distribution of an ensemble of aqueous organic aerosols, evolving via nucleation and concomitant chemical aging. This distribution can be drastically affected by the enthalpy of heterogeneous chemical reactions and the depletion of organic trace gases absorbed by aerosols. A partial differential equation of second order for the temporal evolution of this distribution is obtained from the discrete equation of balance via Taylor series expansions. Once reduced to the canonical …
