Assessment of reconstruction error for laser absorption tomography using large eddy simulation data

Laser absorption tomography (LAT) has been used and validated to produce tomographic data of flow-field properties such as gas temperature and species concentrations. Typically, this has been achieved in exhaust systems and simplified lab-scale flames. A project is ongoing to develop such a system for use with a reacting flow facility at Loughborough University. This facility is designed to given conditions typical of those found in gas turbine (GT) combustion systems, at elevated temperature and pressure using liquid fuels. The application of LAT techniques on this rig will give insight into the more complex flames found in GT systems, increasing understanding of conditions inside the process which may lead to higher emission production. In the long term, this understanding may lead to improved design on combustion systems to reduce emissions. In LAT, path integrated absorbances are measured along laser beams arranged in multiple projection angles. The projection data are then used to reconstruct 2D distributions of temperature and water vapour concentration using inversion algorithms such as Tikhonov regularisation. This paper attempts to quantify the error and uncertainty in the inversion stage by using Computational Fluid Dynamics (CFD) data as a synthetic experiment from which path integrated absorbances can be obtained from the LAT forward problem extracted at multiple positions and times in the simulation. The original cross-sectional planes from the LES simulation are compared with the reconstructed images to assess the error caused by the reconstruction. Methods of comparison include analysis of probability density functions on the full plane or sectionally. Of particular interest is to identify the size and location of major features identified by the absorbance, which will be related to the temperature and species concentrations. The results are used to understand the uncertainty caused by the reconstruction methods. The paper demonstrates the benefit of using CFD to aid the design of new experimental facilities.