Acoustic Localization and Classification of MMOD Impacts on Space Structures Using Deep Learning Networks

Micrometeor and Orbital Debris (MMOD) impacts on spacecraft and large space structures are a significant hazard that can compromise mission success and threaten to endanger the lives of crew. Acoustic emission (AE) signals and impact shock generated by MMOD impacts can be detected by an array of inexpensive, replaceable, wireless surface sensor units affixed to the external surfaces of the spacecraft or space structure. However, due to the complexity of interpreting the AE signals, the determination of impact location and severity of consequent damage is greatly complicated by variations is structure geometry, sensor location, and sensor state. We propose to develop advanced Deep Learning Neural Network (DLNN) classifiers using empirical and model-generated training data to detect the occurrence of MMOD impacts, determine the location of the impact site, and classify the severity of consequent damage. Through accurate estimation of the severity of the damage, appropriate maintenance actions can be performed. Phase I will focus on demonstrating the feasibility of the approach on simple metal structures designed to approximate Whipple shielding and low velocity impacts. Phase II will more fully develop the approach and extend it to more complex geometries, composite materials, and hypervelocity impacts.