Currently, the CBM methods are determined based on expert experience to obtain failure thresholds. The simplistic perfect maintenance measure is utilized to achieve a perfect state. To address these limitations, a CBM method under discrete-state conditions is proposed for multi-component systems. A maintenance list is established based on the historical maintenance data, which contains various maintenance measures. Markov chain is employed to construct a single-component state transition model, which can calculate the component state transition matrix, and obtain the component state recovery matrix. A single-step model is established to predict the state transition probability for each component under different states and different times. The maintenance optimization model is applied to determine the overall maintenance plan of the multi-component system. The proposed method is illustrated using an example involving critical equipment of a subsea production system.