One-step SART with optimized Gaussian filter and artifact removal for dynamic LAS tomography

Laser absorption spectroscopy (LAS) tomography is widely used in combustion diagnosis for gas temperature and concentration imaging. However, limited available projections in dynamic LAS tomography lead to low reconstruction accuracy and severe artifacts. A one-step simultaneous algebraic reconstruction technique (SART) with optimized Gaussian filter and artifact removal was proposed. The reconstructed image after multiple iterations was directly obtained by one-step image reconstruction. A proof of the convergence of SART was developed. An optimization criterion of the Gaussian filter and an artifact removal strategy were derived from the convergence analysis combining with smoothness, continuity, and non-negativity constraints. Numerical and experiment verifications were conducted. Temperature and H₂O mole fraction images of an acoustically excited Bunsen flame and the exhaust flows at a two-stage combustor exit were reconstructed. Both verifications proved good performance of the proposed method with high reconstruction accuracy, few artifacts, and short computation time for reconstruction. In the imaging of exhaust flows at the combustor exit, the average temperature deviation of the proposed method relative to thermocouple readings was 2.08 % when the combustion stabilized, and the average computation time of proposed method was only 0.05 % of the classical SART. The accurate reconstruction without the interference of artifacts enabled the proposed method to provide reliable real-time diagnosis of combustion uniformity for combustor design and optimization.