Condition-based maintenance (CBM) focuses on specific system operation condition to evaluate the real status of monitored equipment. CBM overcomes the blindness of maintenance, and is able to effectively reduce accidental failures caused by insufficient maintenance and waste of resources by the excess. This paper proposes a multi-unit system non-periodic preventive maintenance plan optimization method which is based on state. The concept of "opportunity maintenance threshold" is introduced, and the expression of different maintenance mode transition probability is optimized. We combine multiple maintenance work in accordance with the opportunity maintenance threshold and propose the expression of different maintenance mode transition probability and the objective function on the basis of Markov Decision Process Theory. Through the example analysis, the proposed optimization model can decrease the preventive maintenance activities, save maintenance time and funds, and provide reference for system preventive maintenance strategy decision.