Predictive Modeling of Induction-Hardened Depth Based on the Barkhausen Noise Signal

Abstract

A non-destructive verification method was explored in this work using the Barkhausen noise (BN) method for induction hardening depth measurements. The motive was to investigate the correlation between the hardness depth, microstructure, and the Barkhausen noise signal for an induction hardening process. Steel samples of grade C45 were induction-hardened to generate different hardness depths. Two sets of samples were produced in two different induction hardening equipment for generating the model and verification. The produced samples were evaluated by BN measurements followed by destructive verification of the material properties. The results show great potential for the several BN parameters, especially the magnetic voltage sweep slope signal, which has strong correlation with the hardening depth to depth of 4.5 mm. These results were further used to develop a multivariate predictive model to assess the hardness depth to 7 mm, which was validated on an additional dataset that was holdout from the model training.