Nondestructive evaluation of surface hardness in steam turbine blades of 2Cr13 steel using robot-aided micromagnetic testing system

Abstract This paper presents a robot-aided micromagnetic testing system for the non-destructive characterization of steam turbine blade surface hardness. The system integrates a six-axis industrial robot and a tailor-made micromagnetic sensor to acquire magnetic Barkhausen noise and magnetic incremental permeability signals from the blade surfaces with varying curvatures. The performance of single-point models for convex and concave surfaces is assessed, highlighting the limitations in their generalization ability. To overcome this, a mixed modeling approach is proposed, combining all convex and concave surface data, which significantly improves the model’s ability to adapt to different surface curvatures. The mixed model exhibits enhanced prediction accuracy and stability, addressing overfitting issues encountered in single-point models. The results demonstrate the effectiveness of the mixed modeling approach in achieving reliable and accurate hardness predictions, offering a robust solution for complex surface characterization. Future research will explore further improvements by incorporating more surface types and advanced modeling techniques to enhance the system’s performance.