Auto-Processing of Filtered Rayleigh Scattering Images Including Mie and Background Scattering Contributions

Abstract

A comprehensive and efficient post-processing framework is presented for obtaining aerodynamics measurements of three-component velocity, static temperature, and static density using filtered Rayleigh scattering (FRS). This framework contains several steps which culminate in an optimization process for iterative minimization of errors between the experimentally measured FRS spectra and spectra generated by models, providing an inverse solution to determine the local measurement values. Robustness is gained using a multi-level processing technique for successive spatial resolution refinement of measurements. An assessment of the speed and spectral outputs of several published laser Rayleigh scattering models is provided and resulted in the recommendation that Doll's analytic Tenti-S6 model be adopted as the best trade-off in speed and consistency. A distinguishing feature of the approach is the inclusion of methods for interpreting FRS spectra that contain significant scattering contributions from particles and surfaces. The performance of the methods has been quantified for uncertainties using Monte Carlo simulations, while planar results from two experiments are presented for validation and demonstration of robustness, respectively: a benchtop free jet and the exhaust plume of a miniature turbojet engine. Monte Carlo simulations show that at a signal-to-noise ratio (SNR) of 30 dB the approximate uncertainty in scattering angle is 0.3°, 2.5 K for static temperature, 1.5 m/s for velocity, 0.01 kg/m3 for static density, and 0.01 for both Mie and background scattering intensity ratios. The free jet results show differences of 1.5 K for total temperature, 1 m/s for axial velocity, 0.005 kg/m3 for total density, and 200 Pa for total pressure when compared to probe data. All results fall within the 95% confidence bounds of the a priori uncertainties. The small turbojet (JetCat) results are included to show that the processor is robust and accurate even when used with non-ideal data sets from realistic applications such as vitiated exhausts. The JetCat results show differences of 20 K for total temperature, 8 m/s for axial velocity, 0.01 kg/m3 for total density, and 500 Pa for total pressure when compared to probe data. All results except near the edge of the jet plume fall within the 95% confidence bounds of the a priori uncertainties.