Most tunnel engineering projects face significant challenges in accurately estimating construction duration, primarily due to the complex hydrogeological conditions and the involvement of multiple stakeholders. Addressing the limitation of traditional methods in considering the interactions among multiple risk factors, the Bayesian network is employed to analyze the interactions between various risk factors at the construction process level. Meanwhile, the Program Evaluation and Review Technique (PERT) has been improved to achieve the transmission of risk impacts from each construction process to the overall project duration. Thus, a probabilistic estimation method for tunnel construction duration is established. The proposed method is demonstrated through its application to a subsea tunnel project in China as a typical case study, yielding probability distributions and characteristic parameters for different duration values to provide a scientific basis for project schedule risk assessment. The findings indicate that, compared with traditional methods, the present approach can better reflect actual circumstances and maintain the reliability of prediction results even in extreme cases such as severe delays. Furthermore, the results reveal the uncertainty of duration distribution and its correlation with various risk factors, which strongly supports the dynamic management of tunnel construction schedule risk.