Non-destructive Assessment of AC Breakdown Strength in Silicone Rubber via NIR Spectroscopy

This study explores the application of near-infrared spectroscopy combined with chemometric methods for the non-destructive prediction of AC breakdown strength in silicone rubber insulation. The original spectra were preprocessed using Standard Normal Variate, second-order derivative, and Savitzky-Golay smoothing to mitigate baseline drift and high-frequency noise. Subsequently, wavelength selection algorithms, namely Random Frog and Competitive Adaptive Reweighted Sampling, were employed to identify the most informative spectral variables strongly correlated with breakdown performance. A Partial Least Squares Regression model was then developed to establish a robust quantitative relationship between the selected spectral features and the measured breakdown strength. Results demonstrate that wavelength selection significantly enhances model accuracy and generalizability, with the Random Frog algorithm yielding the optimal predictive performance. This approach provides an efficient, reliable, and non-destructive means for assessing the electrical properties of silicone rubber cable accessories, thereby supporting quality assurance and lifecycle management in modern power systems.