Microstructural-based elastic property analysis in polymer nanocomposites via a fully convolutional semantic segmentation and meta-modeling approaches

Silver (Ag) nanoparticles enhance conductivity in composite structures, offering a lightweight solution for solar panels. However, adequate rigidity is essential for withstanding mechanical and thermal stresses. This study investigates the relationship between microstructural features, weight fraction, dispersion pattern, and elastic properties in PA6/Ag nanocomposites, utilizing experimental data, advanced image segmentation techniques, and machine learning approaches. The measured elastic characteristics include Young's modulus ((Formula presented.)), bulk modulus ((Formula presented.)), and shear modulus ((Formula presented.)). To identify dispersion patterns, scanning electron microscopy (SEM) images of impact-fractured surfaces from different compounds were analyzed using a Fully Convolutional Network (FCN) for precise nanoparticle segmentation. The trained FCN model provided particle size metrics, including minimum, maximum, and average values, which were then used to define a particle area distribution index (ADI) for each reinforced compound. A multiple linear regression (MLR) machine learning model was subsequently developed to correlate Ag weight fraction and ADI with elastic properties. SEM image analysis revealed a generally uniform nanoparticle dispersion, with an increasing tendency toward clustering at higher Ag loadings. ADI values decreased with higher Ag content, indicating a greater clustering tendency at higher Ag loadings. The MLR model results showed that elastic properties were directly influenced by both weight fraction and dispersion behavior, with a higher sensitivity to Ag loading than to particle size distribution trend within the matrix. Highlights: A structure–property relationship was established in PA6/Ag nanocomposite. A novel FCN-based approach quantifies microstructural dispersion An area distribution index linked nanoparticle dispersion to elastic traits A predictive meta-model correlated Ag content, ADI, and elastic moduli. Filler content showed a stronger effect on elasticity than the dispersion pattern.