Measurement of hybrid rocket solid fuel regression rate for a slab burner using deep learning

This study presents an imaging-based deep learning tool to measure the fuel regression rate in a 2D slab burner experiment for hybrid rocket fuels. The is designed verify mechanistic models of reacting boundary layer combustion rockets by measurement rates. A DSLR camera with high intensity flash used capture images throughout burn and are then find calculate rate. U-net convolutional neural network architecture explored segment from experimental images. Monte-Carlo Dropout process quantify uncertainty produced network. computed rates compared values other techniques literature show error less than 10%. An oxidizer flux dependency performed shows predictions accurate independent flux, when training set not over-saturated. Training monochrome successful at predicting noise. superior filtering out noise introduced soot, pitting, wax deposition on chamber glass as well flame traditional image processing techniques, such threshold binary conversion spatial filtering. consistently provides low segmentations allow computation fuel.