Hierarchical Corrosion Assessment of Water Pipelines in a Hydroelectric Plant Through Statistical Analysis and Clustering Methods

Corrosion in carbon steel piping systems is a critical issue in hydroelectric power plants, as progressive wall thickness loss may compromise integrity and operational reliability. This study proposes a data-driven framework combining descriptive statistical analysis and distinct clustering methods to map and rank corrosion conditions in water pipeline systems of a hydroelectric power plant in Brazil. A total of 4916 pipe segments from 20 generators (generating units 1 and 2) were evaluated and the feature matrix included wall loss descriptors, measured thickness statistics, minimum allowable thickness, wall integrity indicators, pipe geometry, segment type, and system information. Descriptive analyses revealed heterogeneous corrosion patterns across generating units, segment types, diameters, and inspection points, with localized severe wall loss conditions in specific segments. A principal components analysis was applied to reduce the original feature space while preserving approximately 80% of the cumulative variance, as suggested by the literature. The best clustering solution was obtained using a weighted consensus model based on the Calinski–Harabasz index, resulting in five degradation/integrity profiles that support segment-level corrosion ranking and inspection prioritization. The proposed framework provides a structured basis for integrating inspection data, statistical descriptors, and integrity indicators into maintenance and decision support processes for hydroelectric power plant piping systems.