A semi-supervised method for the characterization of degradation of nuclear power plants steam generators

The digitalization of nuclear power plants, with the rapid growth information technology, opens door to development new methods condition-based maintenance. In this work, a semi-supervised method for characterizing level degradation plant components using measurements collected during operational transients is proposed. It based on fusion selected features extracted from monitored signals. Feature selection formulated as multi-objective optimization problem. objectives are maximization feature monotonicity and trendability, novel measure correlation between values results non-destructive tests performed assess component degradation. Pareto optimal set normalized defined median obtained values. developed applied real data steam generators pressurized water reactors. shown able identify errors comparable those by ad-hoc tests.