Fig. 6.

High-throughput phase-field simulation and machine learning results. a High-throughput simulation results of the breakdown strength for poly(vinylidene fluoride-hexafluoropropylene) (P(VDF-HFP)) nanocomposites filled by 5 vol% nanofillers with different dielectric constant, electrical conductivity, and Young’s modulus. Here, x₁(ε), x₂(σ), and x₃(Y) represent ε_filler/ε_matrix, σ_iller/σ_matrix, and Y_filler/Y_matrix, respectively. b Comparisons of breakdown strengths between the phase-field model and the machine learning prediction, verified by seven new types of fillers as the testing set (not included in the training set, marked as colored solid symbols) and some experimental results (colored open symbols). The experimental measurement for the breakdown strength of Al₂O₃/P(VDF-HFP) nanocomposite is performed in this work. The experimental breakdown strengths for BTO/P(VDF-HFP), BCZT/P(VDF), and MgO/P(VDF) nanocomposites are from the literatures^37–39