Water inrush is one of the main disasters occurring during tunnel construction in complex geological areas: once it happens, it can cause economic losses, casualties, and delay. Based on risk analysis and management, a risk control scheme is proposed as an effective means to control the risk of such a disaster; however, there are some deficiencies in existing research because the impacts of human factors on the risk of water inrush, dynamic changes in risk information during construction, and the diversity of types of water inrush are neglected. To enrich the research results of water inrush risk control and improve the effect of water inrush risk control, we first use the advantages of Bayesian network to analyse risk events, construct a Bayesian network structure chart of water inrush risk during construction, and propose a fuzzy probability risk analysis model for water inrush. The model can quickly track changes in risk information and diagnose the cause of water inrush disasters while providing an early warning thereof. In addition, considering that the diversity of water inrush types leads to differences in water inrush mechanisms, we believe that the formulation of any water inrush risk control scheme must be combined with water inrush mechanism analysis; therefore, we take a nondefect generated water inrush in front of the tunnel as a representative case and analyse the possible mechanism of water inrush through the stability analysis of the water-resisting strata. Then, based on the results of risk analysis and an analysis of the water inrush mechanism, a reasonable risk control scheme for water inrush is derived.