Deep learning-based image processing for analyzing combustion behavior of gel fuel droplets

Abstract

The success of future rocket engines will rely on their adaptability to new generation fuels that syndicate the desirable properties of solid and liquid fuels, allow full controllability of the engine while being environment friendly as well as cost-effective. To this end, the organic gellant laden gel fuels have attracted recent attention. However, their practical implementation is limited by a lack of understanding of their combustion behavior, even at the droplet scale. Understanding their combustion behavior by analyzing experimental images via the conventional image processing methods is limited due to variable conditions, namely, low image resolution at high-frame rates, frame-to-frame variation in light intensity, and low depth-of-field at high-magnification. In this chapter, a deep learning-based image processing method, namely, the holistically nested edge detection is proposed, which adapts hierarchical convolutional neural networks-based training and can address the challenges faced by conventional techniques for analyzing combustion of gel fuel droplets. © 2022 Elsevier Inc. All rights reserved.