Low-redundancy elemental-line prior selection for interpretable LIBS-based classification of relative aging conditions in composite insulators

Composite insulators are subjected to ultraviolet irradiation, contamination, electric-field stress, and other environmental factors during long-term operation. These effects gradually degrade the surface properties of silicone rubber. To support rapid and interpretable condition assessment, this study proposes a low-redundancy elemental-line prior selection method (LR-ELP) for laser-induced breakdown spectroscopy (LIBS), termed LR-ELP. Four representative surface regions with different wetting characteristics were characterized using the STRI/IEC water-spray test and static water contact-angle measurements and assigned to four relative aging levels. LIBS spectra were then acquired by multi-point scanning and processed through spectral preprocessing and pulse averaging. ANOVA F-score, mutual information, and random forest feature importance were integrated to evaluate wavelength discrimination. A cross-resampling stability constraint was introduced to improve selection reproducibility. Elemental emission-line priors of Si, Al, Ca, Na, O, C, Fe, and H were further incorporated to enhance spectroscopic interpretability. Finally, low-redundancy screening was applied to suppress repeated selection of adjacent highly correlated variables. The resulting LR-ELP-Top-100 feature set retained only 100 of the original 8191 wavelengths. Under five-fold cross-validation on the present dataset, LR-ELP-Top-100 achieved a mean Accuracy of 97.50%, a mean Macro-F1 of 97.49%, and a mean Kappa of 96.67%. Under blocked validation based on consecutive acquisition segments, the mean Accuracy was 91.50%, while a core subset of sensitive wavelengths was repeatedly retained across different training blocks. Overall, LR-ELP substantially reduces spectral dimensionality and redundancy while preserving discriminative information and spectroscopic interpretability. It provides a compact and reproducible spectral-line selection strategy for LIBS-based classification of relative aging conditions in composite insulators.