Thermography and Machine Learning Analysis of Corona Aging of Silicone Rubber Nanocomposites

In the present study, the impact of corona aging on different nanofillers added silicone rubber was studied. The surface morphology of silicone rubber nanocomposites upon aging by corona discharges was analyzed using atomic force microscopy (AFM). The intensity of corona aging is categorized into different classes using surface roughness parameters. The boron nitride (BN) / Aluminium trihydrate (ATH) filler added to the silicone rubber matrix increases the thermal conductivity. Laser-induced thermographic images from the rear side of the samples suggest that the high thermal conductive filler added composites resulted in less local hotspot temperature. A derivative-based strategy on thermal images is useful in detecting the localized thermal hotspot amplitude and depth of defect in different silicone rubber nanocomposites. ATH or BN filler added sample shows 4 times less intense hotspots under normal and corona aging conditions compared to the virgin sample. The Haralick texture features extracted from the thermal derivative images of corona-aged silicone rubber samples are useful in identifying the corona-aging level. Several machine learning (ML) algorithms, including Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Artificial Neural Network (ANN) are employed to classify the corona damage level of silicone rubber samples. However, the Tri-layered ANN algorithm shows a maximum accuracy of 96.8 % with fast training time and prediction speed.