Given data on lengths of cracks at a set of times, a differential equation with random parameters is fitted to the data. A posterior estimate for specimen reliability, given experimental data is obtained.
The model is useful in that it provides a very natural way of summarising material properties that have been determined by experiment. The posterior distributions for the hyperparameters may be used for the general modelling of fatigue.
Coalescence turned out to have an important effect in the data which were analysed. For this reason it was decided to investigate methods of modelling this phenomenon.