Artificial neural network-based predictions of the viscoelastic properties of polymer-matrix composites

Today, the use of machine learning, especially through artificial neural networks, significantly speeds up the design of materials exhibiting unique properties. The high predictive and quantifiable capabilities, performance metrics, efficiency, and verifiability make it a powerful tool in Materials Science. Polymer-matrix composites are particularly notable due to their unique chemical and physical properties, in addition to their intriguing mechanical behavior, which originate from the combination of polymers and nanoparticles. These properties make them outstanding materials suitable for a wide range of technological applications, for example surface activation through microfilm formation, among others. This contribution explores an approach based on artificial neural networks to predict the rheological response of polymer matrix composites obtained using different types of salts. We focus on viscosity and frequency-dependent viscoelastic moduli, which are essential to understand the functional and dynamical properties of the composite, and to quantify the adhesion of composite microfilms on surfaces. Our findings reveal that trained algorithms accurately predict these viscoelastic moduli, as validated by comparison with experimental data over a broad range of compositions. This approach offers a versatile framework for studying other film-forming materials beyond polymer composites, thus representing a general methodology with promising perspective applications for research in Materials Science.