The literature contains a little about coalescence of short cracks in steel. Gao et. al. [17] describe the mechanism of coalescence for high temperature bending fatigue. Reference was also found to the modelling of coalescence phenomena in metal matrix structures, plastics and fiber reinforced composites [59]. It is well recognised that damage accumulation due to a population of cracks is an important mechanism of failure in many different materials [29], [31] . Short cracks are less well understood, and less well researched than their longer counterparts. They are more difficult to examine (since they are so small, and there are more of them), and they grow in a more haphazard fashion. In practice longer cracks tend to be involved in growth only failure, which is caused by a single crack. In contrast, mechanisms for dealing with damage accumulation due to short cracks have focused on quantities such as microcrack density and total length of cracks per unit area.
The microcrack problem involves a large family of cracks that interact with one another. It is observed that these grow into each other to form a dominant long crack which causes failure. For this reason the modelling of the coalescence phenomenon is at least as important as the model for growth.
Initial attempts included examining the data with the idea that each crack had a certain probability of growth, and a probability of being involved in coalescence. This is similar to the jump models used by [48] in analysing growth of cracks.
Some headway was made in the context of focusing on coalescence,
and the directed acyclic graph in Figure 
certainly
formed a start of the solution to the problem.