Surface discharge-induced degradation poses a significant threat to the operational reliability of high-voltage insulation systems. This research investigates the dielectric endurance of polyester-based nanocomposites reinforced with seven distinct nano-additives: iron oxide (Fe3O4), copper oxide (CuO), titanium oxide (TiO2), aluminum oxide (Al2O3), silicon dioxide (SiO2), zinc borate (ZnB) and graphene oxide (GO). Specimens were fabricated at 0.5% and 0.75% weight concentrations and subjected to constant AC electrical stress of 4.5 kV at 50 Hz until failure using the first-plane tracking method. To accurately monitor the aging process, a sophisticated signal processing framework involving Hankel-matrix-enhanced Robust Principal Component Analysis (RPCA) was developed to extract high-frequency discharge features from captured leakage current signals. The degradation characteristics were quantified using various statistical metrics, including Kurtosis, RMS and Burst Discharge Index (BDI). Experimental findings demonstrate that the incorporation of nanoparticles significantly extends the time-to-failure compared to neat polyester, although the effectiveness is highly dependent on both additive type and concentration. At 0.5 wt.%, ZnB exhibited the superior performance in delaying carbonized track formation. However, at 0.75 wt.%, Al2O3 emerged as the most effective additive, achieving a maximum endurance time of 31.61 min. In contrast, certain additives like TiO2 showed a performance decline at higher loadings, likely due to nanoparticle agglomeration. The Hankel-RPCA methodology successfully isolated discharge-specific signatures from background noise, establishing a strong correlation between signal features and material failure stages. This study confirms that the synergy between advanced nanomaterial modification and robust signal processing provides an effective diagnostic tool for monitoring insulation health, offering a vital pathway for the designing of high-performance dielectrics for real-world power system applications.
Pınarbaşı et al. (Sun,) studied this question.