Abstract:In the mesoscale, the proportion of the physical short crack stage in the whole fatigue life is usually high. And the expansion rate of the physical short crack is faster than that of the long crack. This phenomenon has caused potential damage to the material. The early hidden damage such as physical short cracks cannot be effectively detected by traditional nondestructive testing technology. Therefore, metal magnetic memory technology, which can detect early hidden damage and abnormal stress concentration, is introduced to explore physical short crack propagation. Aiming at the difficult parameter characterization problem caused by the specificity of physical short crack growth at the mesoscale, the characterization laws and model of physical short crack growth based on magnetic property parameters (MPPs) were conceived based on micro-experiment. The experiment material is Q235 steel. LWD–1000 long-distance microscope and TSC–5M–32 Tester of Stress Concentration were carried out to observe in-situ the evolution process of crack growth and MPPs, respectively. The characterization laws of MPPs were obtained for physical short crack growth at the mesoscale. The experimental results show that when the length is 0～100 μm, the growth rate increases continuously and peaks occur. The normal component Hp(y) changed from –26.875 A/m to 7.25 A/m with a polar jump. The tangential component Hp(x) changed from –17.24 A/m to –68.78 A/m with a significant increase in absolute value. When the length is 100 μm, the short cracks converge to form the main crack, the crack growth rate slows down, Hp(x) and Hp(y) jump sharply and ΔHp peak appears. When the length is 100～1 000 μm, the growth rate of the main crack increases again and increases continuously. The polarity of Hp(y) changes, the absolute value of Hp(x) increases greatly again, and the ΔHp shows an increasing trend. Furthermore, based on the MPPs, the characterization models of short crack growth rate and remaining life at the mesoscale were established, respectively. The verification results show that the maximum relative errors in the model are 6% and 4%, respectively, which provides a new idea for the characterization and evaluation of early hidden damage in engineering practice.