Data-driven insights into reversible shear adhesion in space-like environments

Machine-learning and statistical methods provide powerful tools to uncover hidden structure-property relationships in complex soft-matter systems such as those operating in space and other extreme environments, particularly when experimental parameters are difficult to decouple.

The accumulation of space debris is a growing challenge for sustainable orbital operations and safety. Reversible dry adhesives offer a promising route for robotic manipulation to capture orbiting debris, but adhesive performance under the combined effects of temperature, vacuum, and material variability remains largely unexplored.

We systematically varied and evaluated shear adhesion across a broad parameter space defined by material composition, geometrical descriptors, temperature, pressure, and interfacial roughness in fabric-elastomer composite systems that combine inextensible textile backings with compliant adhesive layers. Experiments were performed in a custom thermal–vacuum chamber that was designed to measure adhesive performance in simulated space-like environments.

Integrating materials characterization (modulus, viscoelasticity, roughness) with statistical and machine-learning methods, including principal component analysis (PCA) and mutual information analysis, we revealed the relative and combined contributions of temperature, material selection, and interfacial topography to adhesion. This data-driven framework provides structure-property relationships that identify the governing parameters for robust, reusable adhesives for complex environments, including space and other extreme conditions.