View in EDS

A Chemical Kinetic Mechanism Reduction Method with Autoencoder for Hydrocarbon Fuels.

Saved in:
Bibliographic Details
Title: A Chemical Kinetic Mechanism Reduction Method with Autoencoder for Hydrocarbon Fuels.
Authors: Xia, Xiaoqiao1,2 (AUTHOR), Gou, Xiaolong1,2 (AUTHOR) simgxl@cqu.edu.cn
Source: Combustion Science & Technology. 2026, Vol. 198 Issue 8, p2305-2328. 24p.
Subject Terms: *Autoencoders, *Combustion kinetics, *Fossil fuels, *Machine learning, *Chemical models
Abstract: A chemical kinetic mechanism reduction method based on autoencoder (CRAE) is proposed to obtain the skeletal mechanism of hydrocarbon fuels. Based on the network architecture of autoencoder, a computational network structure is constructed using the species molar concentration as the input and output layers. The training datasets are generated from numerical simulations of zero-dimensional homogeneous ignition process using detailed chemical kinetic mechanisms under diverse operating conditions, and the autoencoder is optimized by using back-propagation algorithm and gradient descent algorithm. Subsequently, the weight coefficients connecting the input layer to the first hidden layer are calculated to assess the importance of chemical species, which in turn determines whether a species and its associated elementary reactions can be removed. The skeletal mechanisms for methane and n-heptane generated by using the present CRAE method are compared to those produced by the path flux analysis (PFA) method and the detailed mechanism. The comparisons of the ignition delay time, temperature profiles, and the concentration of important species show that with significantly smaller number of species, the skeletal mechanisms generated by the present CRAE method are more accurate than those of PFA in a broad range of initial pressures, temperatures and equivalence ratios. The CRAE method provides a new and effective means for reduction complex combustion kinetic mechanism of hydrocarbon fuels. [ABSTRACT FROM AUTHOR]
Database: Energy & Power Source
Description
Abstract:A chemical kinetic mechanism reduction method based on autoencoder (CRAE) is proposed to obtain the skeletal mechanism of hydrocarbon fuels. Based on the network architecture of autoencoder, a computational network structure is constructed using the species molar concentration as the input and output layers. The training datasets are generated from numerical simulations of zero-dimensional homogeneous ignition process using detailed chemical kinetic mechanisms under diverse operating conditions, and the autoencoder is optimized by using back-propagation algorithm and gradient descent algorithm. Subsequently, the weight coefficients connecting the input layer to the first hidden layer are calculated to assess the importance of chemical species, which in turn determines whether a species and its associated elementary reactions can be removed. The skeletal mechanisms for methane and n-heptane generated by using the present CRAE method are compared to those produced by the path flux analysis (PFA) method and the detailed mechanism. The comparisons of the ignition delay time, temperature profiles, and the concentration of important species show that with significantly smaller number of species, the skeletal mechanisms generated by the present CRAE method are more accurate than those of PFA in a broad range of initial pressures, temperatures and equivalence ratios. The CRAE method provides a new and effective means for reduction complex combustion kinetic mechanism of hydrocarbon fuels. [ABSTRACT FROM AUTHOR]
ISSN:00102202
DOI:10.1080/00102202.2025.2501695