Tomographic Reconstruction for Flow Parameters Based on Extreme Learning Machine

Tomographic laser absorption spectroscopy is a versatile imaging technique in flow and combustion diagnostics. While the linear tomographic methods need a large number of line-of-sight measurements, the nonlinear methods involve solving optimization algorithms and suffer from high computational cost. Taking advantages of easy implementation, fast learning speed and good generalization performance of the extreme learning machine, a simultaneous tomographic algorithm for temperature and species concentration is proposed. Simulation results show that this algorithm can locate the peaks of the multimodal Gaussian flames faithfully under both noiseless and noisy measurement conditions and can simultaneously image the temperature and species concentration distribution of the tomographic field within 1 millisecond.