Electromagnetic methods to improve the nondestructive characterization of induction hardened steels: A statistical modeling approach

The present paper aims to enhance the accuracy of surface hardness and case depth measurement of induction hardened steel parts using statistical modeling methods and electromagnetic nondestructive techniques. In this research, to create various hardened layers with different thicknesses, eight different conditions of induction hardening treatments have been performed on cylindrical AISI 1045 steel samples. The relationships among outputs of magnetic hysteresis loop/eddy current techniques and surface hardness/case depth obtained from micro-hardness profile measurement have been evaluated. In order to enhance the reliability and simplicity and accuracy of estimation, the principal component analysis was utilized before statistical modeling for dimension reduction and extraction of extra information about the process. The results revealed that the proposed statistical modeling method provides high estimation accuracy, whether the given sample is hardened or not, as well as its depth level. The proposed model not only reduced the number of variables included in estimation (from six to one) but also had a higher accuracy compared to the other classic methods.