Lead University: Lehigh University
PI: Shamim Pakzad

Corrosion is a serious issue causing damage in steel powerline transmission towers that can lead to outages. The transmission tower structures constructed using carbon steels are galvanized and periodically painted to control corrosion. Corrosion is experienced in locations with constant moisture and inaccessibility to repaint. As a solution to this problem, the industry has used weathering steels of high strength and high strength low alloy compositions that lead to light weight overhead transmission towers with improved corrosion resistance, eliminating the necessity of protective paints. In the presence of trapped or circulating moisture the connections of weathering steels with no galvanization continuously produce layers of rust acting like bare carbon steels. These layers increase pressure in the bolts leading do eventual pop-out and failure of the connection called pack-out which is a major issue in transmission lines. Currently there is a large number of transmission towers with carbon and weathering steels with corrosion posing serious threats to their bearing capacity. Timely and accurate inspection and repair is essential to avoid outages due to structural failure. In such circumstances manual inspection of large number of towers for corrosion issues is time consuming, not very accurate due to human error and is not safe for the inspector. An automated inspection procedure for corrosion using image processing techniques with aerial or ground based images is a viable option. The suitable features corresponding to different types of corrosion and stages of damages will be extracted from a database of corresponding images created for this purpose. An efficient classification algorithm will be identified and trained to recognize corrosion damage in the partial images of structures obtained from aerial or ground based inspections. The feature generation and corrosion detection algorithms will be validated using simulated inspections at PPL Training center facilities. The possibility of integration of these results into the images of the whole structure will be explored using scale-invariant feature transform (SIFT) algorithm.