Abstract Additive manufacturing (AM) of polymer nanocomposites offers vast potential for creating multi-functional materials with enhanced mechanical, thermal, and electrical properties. However, the nonlinear interdependencies among nanofiller composition, printing conditions, and final performance metrics make traditional empirical optimization inefficient and unpredictable. To overcome these limitations, this study proposes a Physics-Guided Multi-Task Attention Ensemble (PG-MTAE) model that simultaneously predicts and optimizes the tri-domain properties of nanocomposite-enhanced polymers fabricated via Fused Deposition Modeling (FDM). The model integrates material- and process-level descriptors, a feature interaction module for inter-domain dependency learning, and physics-based constraints to ensure physically consistent predictions aligned with established percolation and reinforcement laws. Two benchmark datasets—the Nanocomposites Properties Database and the FDM 3D Printed Composite Material Dataset were harmonized for multi-domain modeling. The implementation was carried out in Python 3.10. Explainability was achieved using SHAP analysis, while Bayesian Optimization was employed to discover Pareto-optimal configurations for maximizing performance trade-offs. The proposed PG-MTAE achieved a determination coefficient (R 2 ) of 0.9897, RMSE of 0.0348, and MAE of 0.0219, outperforming traditional ANN, RNN, and hybrid XGBoost models by jointly predicts mechanical, thermal, and electrical properties. Results confirm the model’s ability to capture physics-consistent nonlinear interactions among filler concentration, print energy density, and process temperature. This AI-driven approach provides a virtual material design framework, significantly reducing experimental trial-and-error cycles while enhancing design reliability for high-performance, multifunctional nanocomposites in aerospace, biomedical, and electronic applications.
Lakshmaiya et al. (2026) studied this question.