Abstract： In order to ensure the reliability of cluster tools in semiconductor wafer fabrications, a Markov processbased preventive maintenance policy was presented considering the economical efficiency. First, a state transfer model was established by applying the method of combining major maintenance and minimal maintenance,  taking into account the random failure and degradation failure modes. Then, the steadystate probability for each state and transition probabilities among states were calculated based on the Markov theory. After that, the optimal maintenance policy was obtained through the policy iteration algorithm and the Markov decision theory. Finally，numerous experiments were  conducted. The results indicate that the proposed preventive maintenance policy is effective and feasible.