Predicting the deformability of natural short-fiber reinforced polymer composites through combined constitutive mathematical and AI-based modeling approaches

Biocomposites are increasingly used to reduce the use of harmful fossil plastics. Modeling and simulation tools (digital twins) have been developed to replace expensive and time-consuming physical testing during product development based on these materials. In addition to manufacturing aspects and experimentation, this study proposes a micromechanically-based constitutive mathematical model to investigate the viscoelastic–plastic deformability of biocomposites consisting of a polymer matrix and short plant fibers. Due to random fiber orientation and strong bonding between the fibers and the amorphous and crystalline phases of the polymer matrix, influence of the lattice crystalline structure was suppressed. This enables the development of a compact constitutive model. However, constitutive mathematical modeling is computationally time-consuming when applied to predict the long-term deformation behavior in large design spaces. Therefore, the proposed model is used solely to generate high-quality data for machine learning (ML) which is highly computationally efficient. The scaled-up design of new biodegradable polymeric materials, traditionally reliant on costly and time-intensive experimental procedures, is then accelerated by an advanced modeling framework that integrates constitutive mathematical models with AI-based approaches.