A Binary Valued Reconstruction Algorithm for Discrete TDLAS Tomography of Dynamic Flames.
Shuang Qiu,Zhang Cao,Jinting Wen,Xiaoqian Zhang,Lijun Xu
DOI: https://doi.org/10.1109/tim.2023.3271738
IF: 5.6
2023-01-01
IEEE Transactions on Instrumentation and Measurement
Abstract:A discrete tomographic imaging method is introduced to reconstruct distributions of temperature and gas molar concentration in tunable diode laser absorption spectroscopy (TDLAS). The proposed method uses projection histograms along laser paths from different projecting angles. Continuously distributed temperature and specific gaseous concentration in the combustion region of interest are discretized into two finite sets of value pairs in the feasible ranges of temperature and gas concentration, and in this way, these two distributions are well approximated by the value discretization. The Voigt model between absorption spectral profiles and gas temperature, as well as concentration, is simplified to a linear matrix equation, and its binary-valued solution yields values of temperature and gaseous concentration from these two finite sets. These sets are composed of discrete value pairs of the gas temperature and molar concentration. The projection histogram along each laser path is determined by the coefficients of the gas temperature and concentrations from the measured line-of-sight absorption spectrum. From projection histograms along various laser paths, cross-sectional distributions of temperature and concentration are simultaneously obtained from the combinations of distributions of optimally selected discrete pairs. Numerical simulations are conducted to evaluate the proposed discrete tomography method, and relative errors of temperature and concentration are reduced by more than 30% compared with the classical two-color method, especially in noisy cases. A McKenna burner at different airflow rates and Bunsen flames with acoustical excitations are applied to verify the proposed method in stable and dynamical flames. Both simulated and experimental results show that the proposed method generates better images and noise-immune performance.