This paper deals with the ordering policy of spare parts in multi-unit systems, which is affected by various factors. In the investigation, first, the demand of preventive maintenance for spare parts is divided into two parts, one for the units with random failures and the other for the units with hidden faults. Then, by using the probability theory, a spare part demand model of exponential life-type units is established for the equivalent-cycle preventive maintenance. Based on the proposed model, two kinds of ordering policies of spare parts, respectively with initial inventory and ordering lead time, and their Visual Fortran simulation optimization algorithm, are presented, with the total cost of order and storage in service age as the optimization goal, and with the ordering times as the optimization variable. A case study is finally performed to calculate the economical order times, the ordering policy and the corresponding total cost of order and storage. It is found that the ordering policy of spare parts is periodical, and that the optimized ordering policy, which considers the initial inventory, the ordering lead time, the ordering cost and the unit storage cost of spare parts, provides an effective reference for the enterprises with different initial inventories of spare parts to determine a suitable ordering policy.