Towards an active-learning approach to resource allocation for population-based damage prognosis