A machine learning based model for gray gas emissivity and absorptivity of H 2 O-CO 2 -CO-N 2 mixtures

Wei Chen,Tao Ren,Changying Zhao
DOI: https://doi.org/10.1016/j.jqsrt.2023.108798
IF: 2.342
2024-01-01
Journal of Quantitative Spectroscopy and Radiative Transfer
Abstract:To quickly estimate the radiative transfer of gases, it is common to adapt parameters such as total (gray) emissivity and absorptivity for the calculations. However, existing state-of-the-art models for gray gas mixtures rely on two-dimensional look-up tables and multiple approximation steps, which can result in reduced accuracy and efficiency. This study addresses these limitations by employing a multilayer perceptron neural network approach to develop accurate and efficient absorptivity calculation models. Remarkably, since absorptivity equals emissivity when the absorbing gases and emitting wall surface have the same temperature, a single model can predict both quantities accurately. The training datasets were calculated based on line-by-line integration from HITEMP-2010. Two models were provided based on different spectral line-shape models (pseudo-Lorentz and Alberti cut-off), both of which take care of the strong line mixing effects at higher pressures. To enhance user accessibility and usability, we developed a web-based application that serves as a graphical user interface for the models. The application is conveniently accessible via a web browser.
optics,spectroscopy
What problem does this paper attempt to address?