Nowadays, the main challenge in maintenance is to establish a dynamic maintenance strategy to significantly track and improve the performance measures of multi-state systems in terms of production, quality, security and even the environment. This paper presents a quantitative approach based on Dynamic Bayesian Network (DBN) to model and evaluate the maintenance of multi-state system and their functional dependencies. According to transition relationships between the system states modeled by the Markov process, a DBN model is established. The objective is to evaluate the reliability and the availability of the system with taking into account the impact of maintenance strategies (perfect repair and imperfect repair). Using the proposed approach, the dynamic probabilities of system states can be determined and the subsystems contributing to system failure can also be identified. A practical application is demonstrated by a case study of a blower system. Through the result of the diagnostic inference, to improve the performances of the blower, the critical components C, F, W, and P should be given more attention. The results indicate also that the perfect repair strategy can improve significantly the performances of the blower, while the imperfect repair strategy cannot degrade the performances in comparison to the perfect repair strategy. These results show the effectiveness of this approach in the context of a predictive evaluation process and in providing the opportunity to evaluate the impact of the choices made on the future measurement of systems performances. Finally, through diagnostic analysis, intervention management and maintenance planning are managed efficiently and optimally.