Concentration Imaging of Multiple Gases from H2O Absorption Spectrum Coupled with CFD Model

Tunable diode laser absorption spectroscopy (TDLAS) tomography is a well-proven combustion diagnosis method, but has difficulty especially in the simultaneous imaging of multi-component concentration. In this work, a prediction method was proposed for concentration imaging of multiple gases from H2O absorption spectrum coupled with the computational fluid dynamics (CFD) model. CFD simulations of methane/air Bunsen burner flames were implemented to reveal the relationship among mole fractions of multiple gases based on the chemical reaction mechanism in the combustion process. The distributions of H2O, CO2, and CO mole fractions at different equivalence ratios and axial heights were then obtained as references. A back propagation (BP) neural network was trained to predict projection values of CO2 and CO mole fractions from integral absorbances of H2O spectrum centered at 7185.6 cm(-1) and 7444.4 cm(-1) along the same projections. And the datasets for network training and testing were generated using a pentagonal fan beam arrangement derived from a real TDLAS tomographic sensor. Eventually, distributions of flame temperature as well as H2O, CO2, and CO mole fractions were reconstructed by using the measured integral absorbances and predicted mole fraction projection values. Reconstructed images at different noise levels in the numerical simulation showed high structural similarities with the reference images provided by CFD models.