Maintenance management has a direct influence on production performance. Existing works have not systematically taken the on-line production information into consideration in determining maintenance work-order priority, which is often assigned either through an ad hoc approach or using largely heuristic and static methods. In this paper, we first present a metric that can be used to quantitatively evaluate the effects of different maintenance priorities. Based on this index, one can employ a search algorithm to obtain maintenance work-order priorities that will lead to improved productivity within the optimization horizon. These concepts and methods are validated through simulation experiments and implementation in a real industrial facility. The results show that the effective utilization of on-line production data in dynamic maintenance scheduling can yield visible production benefit through maintenance priority optimization.