ABSTRACT
EDF periodically performs in-service inspections of passive components within its electric power plants in order to ensure that their degradation is lower than a critical level and to guarantee the safety and the availability of the installations. These examinations allow to collect successive degradation measurements data (crack sizes) for the components. Unfortunately, these data are incomplete. First, small cracks with size below a specific known threshold may be detected without possible measurement. In that case, the only available information hence is: presence of a crack with size below the threshold. Secondly, one single measurement can be performed by an examination and, in case of several competing cracks on one component, the process cannot measure all cracks sizes but only the largest one (which is not necessarily the same throughout the whole component lifetime) if its size exceeds the previous threshold. However, even if they are not measured, all cracks are detected, so that, in that case, the available information is: number of cracks, size of the largest one. Taking into account this partial information, a specific stochastic model is proposed. In this model, cracks initiate following a Poisson process and propagate according to gamma processes. Parametric estimation procedures are developped, tested on simulated data and then applied to the industrial data. The fitted model is next used to make some prediction over the future degradation propagation and over the residual operation time upon which a critical degradation level is reached.
