Failure of metal structures is caused by cracks appearing and growing in the material until the strength of the structure is compromised. The way in which such cracks grow in metal has been researched extensively; the great majority of this work concentrates on long cracks, that is cracks of the order of 1mm or longer. Short crack propagation is affected by the microstructural properties of the material, and differs significantly from the growth of long cracks. These short cracks grow into each other to form longer cracks. This is an important mechanism for damage accumulation, and is termed coalescence.
A model is developed for short crack propagation which takes into account the microstructural features of the material. This gives reasonable reliability predictions when coalescence is not a dominant feature. A further model is developed for coalescence. These two models are then combined, using auxiliary variables for unobserved parameters of interest.
Data on the growth an coalescence of short cracks has been provided
from experimental tests carried out at the Department of Mechanical and
Manufacturing Engineering, Trinity College Dublin. The analysis is carried
through in a Bayesian framework. Markov Chain Monte Carlo (MCMC) techniques
are used to sample from the posterior distribution of the parameters of
the model. Recommendations are made as to the nature of the spatial data
that may be easily recorded in the future, in order that reliability predictions
may be improved.