Abstract: Aiming at the problem that stress concentration could occur during plastic deformation or service of materials, based on the relationship between the characteristic parameters extracted by 3MA comprehensive nondestructive testing technology and the surface hardness of pipeline steel, the stepwise regression model and BP neural network model were constructed to realize the detection of the surface hardness of pipeline steel. The results show that there was a correlation between electromagnetic characteristic signal and surface hardness. Whether stepwise regression model or BP neural network model, in the 10% error range, the confidence could reach 100%.