Surface heat treatment to improve the metallic hardness via electric arc energy

This study investigates hardness evolution in cylindrical S45C steel subjected to tungsten inner gas (TIG) arc surface hardening and develops a predictive framework based on artificial neural networks (ANN). Hardness measurements obtained from 25 specimens at five subsurface depths ranging from 0.2 to 1.0 mm, yielding approximately 2000 data points, were used to train and validate the model. The ANN shows high predictive capability, with predicted hardness closely matching experimental results and the highest reliability observed within the functional hardened region at depths of 0.6–0.8 mm. Process analysis indicates that heat input is primarily governed by current intensity and rotational speed, while axial speed mainly controls hardening track overlap, with current intensity identified as the dominant factor influencing hardness penetration depth. Microstructural characterization reveals the formation of a martensite-dominated hardened layer containing bainite and retained austenite, followed by a heat-affected zone and a ferrite–pearlite base structure with grain coarsening, while microhardness profiles confirm the development of a hardened layer extending to approximately 1.5 mm, suitable for practical engineering applications. The developed ANN model successfully represents process–structure–property relationships in TIG arc hardening and enables reliable prediction of hardness distribution for process optimization in industrial surface treatment applications.