Experimental analysis and machine learning-based recognition of multiscale solid particles moving in gas

Abstract

The rapid growth of artificial objects in Earth orbit poses increasing risks to satellites, human missions, and the long-term sustainability of space operations. Uncontrolled collisions with such fragments can damage spacecraft, threaten human missions, and jeopardize the long-term safety of spaceflight operations. Detecting and characterizing small debris—particularly in the sub-millimeter to millimeter range—remains a major challenge as far as small particles size lies below the effective resolution of most ground-based radar and optical systems. To address this, we present an integrated experimental–computational framework that combines high-speed flow visualization with machine learning–based recognition. Shock tube experiments at rarified air flow velocities from 50 up to 900 m/s were used to reproduce conditions representative of particle motion. Particle dynamics were recorded using high-speed shadowgraphy and processed through two complementary pipelines: a contour-based algorithm extracting geometric parameters (width, height, area) together with average brightness relative to the background, and a YOLOv11 deep learning model trained on annotated datasets for real-time particle detection and trajectory reconstruction. The method enables automated generation of particle size distributions, brightness statistics, x–t diagrams, and velocity–time profiles for particles in the 10 μm–5 mm range. The novelty of this work lies in combining high-speed laboratory shadowgraphy with both classical computer vision and deep learning methods, enabling simultaneous extraction of geometric, optical, and kinematic particle characteristics with direct relevance to orbital debris studies. By linking controlled laboratory experiments with scalable computer vision tools, this approach provides a basis for studying fine-particle debris dynamics, validating flow models, and advancing space situational awareness capabilities for debris detection and mitigation.