Machine Learning Assessment of Fiber Morphology Descriptors in Solution Blow Spinning

Rivera Vera, Claudia Ivette; Cid Galiot, Jonathan Josué; Grande Ramírez, José Roberto; Domínguez Herrera, José Ernesto

doi:10.22034/jams.2026.260402

Machine Learning Assessment of Fiber Morphology Descriptors in Solution Blow Spinning

Document Type : Original Article

Authors

Claudia Ivette Rivera Vera ¹

Jonathan Josué Cid Galiot ²

José Roberto Grande Ramírez ²

José Ernesto Domínguez Herrera ²

¹ Departamento de Nanotecnología, Universidad Tecnológica del Centro de Veracruz, Cuitláhuac, Veracruz, México

² Coordinación de posgrado, Universidad Tecnológica del Centro de Veracruz, Cuitláhuac, Veracruz, México

10.22034/jams.2026.260402

Abstract

Machine-learning-assisted reanalysis of two previously published Box–Behnken datasets for solution-blow-spun (SBS) polystyrene (PS) and polysulfone (PSf) mats is presented. The analysis focused on fiber diameter (DoF) and the relative amount of fibers (RoF). Regression models were trained using coded processing variables and evaluated by 5-fold cross-validation. For PS, whose morphology was predominantly fibrous, DoF showed the strongest predictive performance; linear regression achieved a coefficient of determination (R²) of 0.8142, a root mean square error (RMSE) of 142.45 nm, and a mean absolute error (MAE) of 91.32 nm. In contrast, PSf mats, characterized by bead-on-string morphology, showed substantially lower predictive performance for DoF and RoF, with mean DoF reaching an R² of only 0.0043. These findings indicate that descriptor suitability depends on the structural regime of SBS materials. Fiber diameter is an informative characterization parameter when continuous fibers dominate the architecture but becomes insufficient when bead-related features contribute substantially to the morphology. Machine learning can therefore complement conventional morphological characterization by helping identify which descriptors most effectively represent processing–structure relationships in solution-blow-spun fibrous materials.

Keywords

Fiber diameter

Machine learning

Morphology prediction

Solution blow spinning