This paper describes the development of an original eddy current method for the characterization and classification of different aging states of heat resistant austenitic steel tubes, commonly used in petrochemical industry to produce oil derivatives. These tubes are exposed to high temperatures causing microstructural transforma-tions. They are also under oxidizing environments leading to the formation of an external surface with ferromagnetic behavior. An eddy current testing (with a Hall sensor) was used in order to observe magnetic changes in the specimen. The amplitude and phase-shift of the eddy current signals are calculated and used as features for the samples characterization. An electromagnet was implemented in order to overpass the ferromagnetic external surface and measure the base metal response. A finite element simulation was also developed in order to estimate the skin depth of the eddy currents in samples with different aging states. A machine learning algorithm has been used to classify the test specimen based on the extracted features. Results suggest that the proposed method is a potential non-destructive technique for the characterization and classification of heat-resistant austenitic steel tubes with different aging states.