Previous studies seldom consider the difference of preventive maintenance effect on different failure processes when establishing the preventive maintenance model. Meanwhile, preventive maintenance models usually assume that minimal repair is performed for each failure, but this assumption is not applicable in some systems. Once this assumption is made for the inapplicable system, the maintenance strategy obtained by the established preventive maintenance model will not be optimal. In order to get a better maintenance policy to reduce maintenance cost, this paper divides the system failure processes into two types: wear-out type and damage-type. The model assumes that preventive maintenance can reduce the probability of damage-type failure, but it has no effect on wear-out type failure. At the same time, for the system studied in this paper, it is assumed that after corrective maintenance, the overall state of the system is worse than the moment just prior to the failure. Then, an extended preventive maintenance optimization model based on generalized Polya process is established. Finally, a practical algorithm is designed to solve the model. The analysis of numerical examples shows the potential value of the proposed method, which may provide maintenance guidance for enterprises and save maintenance costs.