AI-Arrester-Diagnostics

October 21, 2023

In my thesis, I aimed to automate the diagnostic process of lightning arresters. The existing manual method was labor-intensive, requiring the manual extraction of the arresters' temperatures . To simplify this, I introduced a new method using Artificial Intelligence Deep Learning. I employed a Mask R-CNN model to segment lightning rods from thermal images, which was then followed by an algorithm to extract temperature data for degradation diagnosis. The automated method proved to be a feasible alternative to the manual process, showing promise in improving the efficiency and accuracy of lightning arrester inspections. Future work will involve refining the segmentation process and further tuning the model for better accuracy.