A method for allocating resources to machines of a production facility comprises: receiving forecast data indicating planned demands for the resources and a demand deviation for each resource; generating a new demand for each resource from the planned demand and the demand deviation of the resource in several iterations; assigning capacities to the new demands in each iteration by determining resource-machine combinations, wherein a priority is determined for each resource and each machine based on a set of priority rules, wherein the resource-machine combinations are determined by combining the resources and the machines according to the priorities; and generating a roll-out plan assigning resource-machine combinations to be rolled out to future time periods, wherein the roll-out plan is generated from the resource-machine combinations of different iterations and an estimated roll-out time for each resource-machine combination such that a total number of the resource-machine combinations to be rolled-out is minimized.

A resource allocation system for a production facility stores: a production schedule assigning production tasks to machines of the production facility, each production task comprising a production time window during which a machine is occupied with the production task; a maintenance schedule assigning maintenance tasks to service resources, each maintenance task comprising a maintenance time window during which a machine is occupied with the maintenance task; and a maintenance plan comprising maintenance rules for generating maintenance tasks, the maintenance rules at least comprising a scheduling rule from which a time after which a machine has to be maintained is derivable. The resource allocation system is configured for determining maintenance tasks for machines based on the maintenance rules, wherein the time window of a maintenance task for a machine is determined from a previous maintenance task, the scheduling rule for the machine and the production tasks for the machine, such that the time window of the maintenance task does not overlap with the production tasks.