Numerical demonstration of 3D reduced order tomographic flame diagnostics without angle calibration

Abstract

Towards a calibration-free 3D flame measurement, this paper demonstrates a reduced order tomographic reconstruction method without the need to calibrate view angles. First, time varying 3D flame structures and two-dimensional (2D) view projections were generated by computational fluid dynamics (CFD) and ray-tracing simulations. Second, 3D flame and 2D projections were used as the training data with model reduction operation, and a mapping relationship was trained between the reduced order data. In this manner, a reduced-order tomography algorithm without any explicit information of view angle was realized. Third, 2D projections with view angles different from the training data were then used to demonstrate the validity of the algorithm. The results with undetermined arbitrary view angles presented a robust reconstruction as evaluated by reconstruction error and peak signal-to-noise ratio, while the camera optical center deviation effect can be fairly eliminated with projection realignment according to the flame centroid. The results of this work are expected to reduce operation difficulty and promote the applicability of multi-angular tomographic diagnostics of turbulent flames.