Data mining algorithm for solid propellant micro-CT based on computer vision techniques

Solid propellants exhibit intricate internal microstructures and heterogeneous component distributions, posing significant challenges for data analysis and information extraction based on Micro-CT imaging. To address these difficulties, this study proposes a high-efficiency, high-accuracy image analysis framework designed to deeply mine Micro-CT data of solid propellants and facilitate the elucidation of the correlation mechanisms between microstructural features and macroscopic performance. High-resolution Micro-CT scanning was conducted on two representative propellant samples. Through a series of key processing steps, including image pre processing, component segmentation, density estimation, and particle size analysis, the framework enables the quantitative extraction of critical microstructural parameters，such as overall density，the particle size distribution of ammonium perchlorate(AP), and the spatial distribution characteristics of each constituent component. This process achieves accurate identification of internal phases, yielding density estimates with errors ≤ 5% and particle size evaluations with errors ≤ 8%.Furthermore,the study quantitatively characterized the centroid distributions and radial spatial distributions of propellant components, thereby enabling the assessment of their dispersion and agglomeration states within the matrix. These results provide robust data support for the optimization of propellant formulations, investigation of interfacial bonding mechanisms, and evaluation of storage life. The proposed approach thus offers a scientific foundation for a deeper understanding of the structure-property relationships governing solid propellant behavior.