Hyperspectral Imaging and Analysis for Steel Paint Condition Assessment (SN-9)

This project aims to develop a time-efficient, safe, and reliable technology based on hyperspectral imaging to inspect the service condition of anti-corrosion paint/coating on steel structural members (e.g., steel girders). Specific objectives are to establish a database of paint/coating degradation and develop a supervised machine learning based classification tool for inspection of steel coatings. The objectives will be achieved through a holistic literature and market survey, a comprehensive lab-based experimental study, and a machine learning modeling. The survey will determine representative paints and their major degradation mechanisms, which will be used to guide experiment design. In the lab, representative paints will be collected, applied on steel members, and degraded under accelerative conditions (e.g., enhanced ultraviolet, heat, moisture, and temperature/humidity swings). The paints, aged to various extents, will be rated qualitatively according to standard methods and characterized quantitatively to output their degrees of degradation. Hyperspectral imaging will also be used to characterize the aged paints to generate spectral features, which will be correlated to the qualitative and quantitative evaluations of the paint conditions. All the results will form a database, in which the hyperspectral features will be used as input parameters and the paint condition parameters will be used as output parameters to train a machine learning-based classifier. The classifier will be validated so to serve as an inspection tool for paint condition on steel structural members, as well as to assist in decision-making for maintenance protocols