Abstract: The short fatigue cracks of metal materials are systematically sorted out in terms of their definition, classification, test methods, initiation mechanism, and typical growth rate models, as well as the problems facing the current research on short crack behavior, are summarized and prospected. The results show that the microstructure of materials plays an important role in the initiation and propagation of short cracks. When the crack tip reaches the grain boundary, the growth rate of cracks decelerates due to the influence of grain orientation and grain boundary resistance, and the short crack propagation path appears to be locally deflected after the crack tip breaks through the grain boundary constraint. In terms of the short cracks propagation behavior description, the driving force of short crack propagation is related to the load level, residual stress, surface treatment and local wear condition, and so on. During propagation, the crack tip materials yield in a small area and produce a local plastic zone. At the same time, the plastic wake is gradually formed, which results in plastic-induced closure and a decrease in the crack propagation threshold value. Besides,the interaction between crystal dislocations and dislocation stacking can hinder the dislocation development. The machine learning algorithm, as an advanced technological tool, has been applied in the characterization of short crack growth behavior, which has contributed to the improvement of prediction accuracy. With further research on the behavior of short cracks in metal materials, we can focus on real-time monitoring of short crack initiation and propagation, control of test influencing factors, discrete data analysis,and engineering applications of the growth rate model in the future. The establishment of a unified characterization model for long and short crack growth rate as well as real-time monitoring and safety assessment techniques will be explored. In addition, more effective and accurate safety assessment and remaining life management of critical metal structures can be realized by combining advanced machine learning algorithms.

Key words: metallic materials, short fatigue crack, initiation mechanism, growth rate model, machine learning