A new magnetic data-driven approach for the assessment of microstructural properties of steel components in service is provided by deep learning modeling.
Based on the interrelated characteristics between the microstructure and magnetic properties of ferromagnetic materials, a novel hardness prediction model based on generalized regression neural network (GRNN) and magnetic measurements are proposed in this paper.
Using the magnetic properties as training and testing sets, the GRNN-based model achieves high-precision estimation of the hardness of 35CrMo steel, which is significantly better than the linear regression method.
With the intrinsic relationship between the hardness and magnetic coercivity of 35CrMo steel, a hardness prediction model based on GRNN and magnetic data is proposed; the predictions of hardness are generally in line with the actual values, with an error of 1.18%.
