I. Ahadi Akhlaghi - Department of Electrical and Bioelectric Engineering, Sadjad University of Technology, Mashhad, Iran
S. Kahrobaee - Department of Mechanical and Materials Engineering, Sadjad University of Technology, Mashhad, Iran
M. Sekhavat - Department of Mechanical and Materials Engineering, Sadjad University of Technology, Mashhad, Iran
H. Norouzi Sahraei - Center of Nondestructive Evaluation (CNDE), Sadjad University of Technology, Mashhad, Iran
F. Akhlaghi Modiri - Center of Nondestructive Evaluation (CNDE), Sadjad University of Technology, Mashhad, Iran

The present paper shows the results of applying an artificial neural network to three non-destructive magnetic methods including magnetic hysteresis loop (MHL), eddy current (EC), and magnetic flux leakage (MFL) techniques to determine mechanical features of plain carbon steels with unknown carbon contents subjected to tempering treatment. To simultaneously evaluate the effects of carbon content and microstructure on the magnetic and mechanical properties, four grades of hypoeutectoid steel samples containing ۰.۳۰, ۰.۴۶, ۰.۵۴, and ۰.۷۱ wt.% carbon were austenitized in the range of ۸۳۰-۹۲۵ °C and then subjected to quench-tempering treatments at ۲۰۰, ۳۰۰, ۴۰۰, ۵۰۰ and ۶۰۰ °C. In the next step, mechanical properties including tensile strength, elongation, and hardness were measured using tensile and hardness tests, respectively. Finally, to study the electromagnetic parameters, MHL, MFL and EC non-destructive electromagnetic tests were applied to the heat-treated samples and their outputs were fed to a generalized neural network designed in this work. The results revealed that using a proper combination of electromagnetic parameters as the ANN input for each mechanical parameter enables us to determine the hardness, UTS and elongation of hypoeutectic carbon steel parts after tempering treatment with high accuracy.

mechanical parameters, Magnetic hysteresis loop, Magnetic Flux Leakage, Eddy current, Artificial Neural Network