Analysis of Noisy Radiometer Data from ISS Reduced Gravity Droplet Combustion Experiments

Abstract

Reduced gravity droplet combustion experiments were performed on the International Space Station (ISS). The data from these droplet combustion experiments have been made publicly available by NASA. These experiments utilized narrow-band and wide-band radiometers to measure thermal radiation heat fluxes from flames. The initial set of radiometers, which were used in 80 experiments, exhibited large noise levels, making interpretation of data from these experiments difficult. The noisy radiometers were replaced by NASA with a second set of radiometers that had significantly less noise. The work presented here is focused on: (1) evaluating methods to “denoise” the data from the initial (noisy) set of radiometers with a goal of detecting flame oscillations; and (2) making these denoising methods available to other researchers who would like to analyze the noisy radiometer data. The denoising efforts employed local polynomial regression analysis with a Gaussian kernel as well as frequency-based low-pass filters. Windowed Fourier transform and wavelet transform approaches were also employed to detect flame oscillations in radiometer data. Comparison of results from these approaches showed that local polynomial analyses and also wavelet transforms can detect the presence of flame oscillations in noisy radiometer data. Uncertainties in the noisy radiometer data were evaluated using a Monte Carlo bootstrapping approach.