The PMC project advocates a non-confrontational cooperation model. As an extension of the owner, the PMC contractor manages the project in all aspects and processes on behalf of the owner, and the level of trust between the two parties determines effectiveness of management, which in turn affects the success or failure of the project. This paper adopted a dynamic Bayesian network approach, based on social exchange theory, to establish a dynamic model of the evolution of trust level between owners and PMC contractors, and used the two-way inference function of dynamic Bayesian network to identify the key indicators and dynamic evolution law of trust level. The results of the study indicated that the trust level between the owner and the PMC contractor depends mainly on the level of calculative and relational trust, and that the most important drivers of calculative and relational trust are the reciprocal behaviors and contractual flexibility. In addition, the reciprocal behaviors has a greater driving effect on the promotion of relational trust, and the contract flexibility has a greater driving effect on the promotion of calculative trust. Also, the level of calculative trust is higher in the early stage of project implementation, and the more calculative trust is generated, the more relational trust is generated in the middle and later stages. In the middle and later stages of the project, the importance of relational trust gradually increases and surpasses that of calculative trust, and the decline of relational trust has a greater negative impact on the overall trust level than calculative trust.