Advanced Techniques for Assessing Aging Distribution in Silicone Rubber Composite Insulators: A Micro to Macro Perspective

Composite insulators are increasingly replacing traditional insulators in the large-scale construction of power grids. However, these insulators are susceptible to aging under outdoor conditions, which can lead to changes in the composite materials and flashovers, resulting in power outages and economic losses. Detecting the microscopic changes associated with aging in composite materials is challenging. This study introduces an unsupervised, pixel-level hyperspectral assessment method for evaluating the aging distribution in composite insulators. Hyperspectral imaging was used to collect data from retired insulators, followed by spectral and spatial data preprocessing to minimize noise. A shadow segmentation index was developed to address image shadows and backgrounds, facilitating the extraction of the target insulators. The method involves clustering spectral and spatial information using mean shift, followed by superpixel segmentation based on entropy rate to distinguish different aging characteristics. Dimensionality reduction and feature extraction were performed using SuperPCA on the hyperspectral data. Finally, Kmedoids clustering was applied to the superpixel features to visualize the aging distribution based on spectral signatures. Comparative analyses of different methods on public datasets, validated by external metrics, demonstrate the superior performance of the proposed unsupervised approach. The application of the proposed method effectively assesses and visualizes the aging distribution of composite insulators, with results that align with the aging patterns. This research provides insights into material aging patterns across various environments and offers guidance for the material and structural design of composite insulators. Furthermore, it has potential applications in the noncontact evaluation of other electrical equipment and composite materials.