Inverse Design of Polyarylate Nanofiber-Enabled Basalt Scale Nanocomposite Insulating Paper for High Dielectric Reliability via Finite Element-Guided Machine Learning.
Hu Qian Q, Wang Jingxian J, Gan Qiwen Q, Liao Ying Y et al.
The dielectric reliability of polymer-based insulating paper critically depends on whether its structure-property relationships can be predicted and programmatically tuned. However, most dielectric papers are still developed through empirical trial-and-error, which limits precise design and efficient optimization toward target performance metrics. Here we establish a digitally enabled inverse design framework that integrates finite element analysis (FEA) with machine learning (ML) to enable the predictable fabrication of high-performance γ-BS/PAR nanofiber composite insulating paper (γ-BS: silane-modified basalt scales; PAR: polyarylate nanofibers). A physics-based structure-property dataset was generated via microstructure-resolved FEA and used to train an ML surrogate that not only predicts breakdown strength but also inversely recommends process/structure parameters within a feasible processing window. Guided by the model, the optimized nanocomposite paper achieves an ultrahigh breakdown strength of 125.63 kV mm-1, markedly exceeding commercial aramid paper. The nanocomposite paper further retains >80% of its breakdown strength after UV/thermal aging and repeated folding and exhibits enhanced flame retardancy and thermoplastic-driven reprocessing potential; despite progressive performance attenuation during recycling, recycled samples maintain breakdown strength superior to commercial aramid paper, benefiting from the thermoplastic re-weldability of the PAR network. "Simulation-to-data-to-manufacturing" paradigm provides a quantitative route for designing polymer dielectrics with controllable performance and scalable processability.