Digital Twin Design of Nanocarbon/Polymer Composites: Physics-Informed Neural Networks Bridging Percolation to Sustainable Manufacturing

Nanocarbon fillers, including graphene and carbon nanotubes, possess exceptional intrinsic properties; however, their full potential in polymer composites remains unrealized due to persistent gaps in predictive modeling and the translation of interfacial chemistry into quantifiable parameters. This critical review synthesizes computational frameworks bridging molecular-scale functionalization to macroscale performance, emphasizing percolation theory, thermal transport, and micromechanics. We systematically evaluate effective medium theories incorporating Kapitza resistance, finite element morphological modeling, reactive force field molecular dynamics, and physics-informed neural networks (PINNs) for real-time dispersion kinetics. Quantitative analysis reveals that interfacial thermal resistance frequently exceeds 60% of total thermal impedance, with functionalization chemistry directly governing computational parameters: covalent silanization reduces Kapitza resistance and elevates critical interfacial shear strength (70–85 MPa), whereas oxidation degrades mechanical coupling (25–35 MPa). Critical bottlenecks are identified, including the MD-to-device scale gap (∼1 nm to ∼1 mm), interlaboratory variance (±30%), and energy-intensive synthesis (∼50 MJ·kg−1). We propose hierarchical interface architectures, generative adversarial networks for morphology prediction, and dynamic covalent vitrimers enabling closed-loop recycling with <5% property drift. A roadmap is established for concurrent multiscale PINNs integrated with in situ characterization and FAIR data practices, targeting carbon-neutral manufacturing with >90% property retention and ±10% predictive accuracy by 2030.