Because the theory of dual prevention mechanism is put forward for a relatively short period of time and lacks the support of corresponding regulations and standards, it is still difficult to see the effectiveness of real and effective operation of dual prevention mechanism to prevent accidents in enterprises, and there are some professional and technical obstacles that have not been overcome. This paper shifts the focus of the establishment of the dual prevention mechanism to focus on safety risk control and develop corresponding emergency measures. The enterprise safety index system is established from the two parts of risk grading and control and hidden danger investigation and management. According to the definition of Bayesian formula network, determine the conditional probability of Bayesian network, and construct the safety risk assessment model based on Bayesian network. The risk reachable probability of the model constructed in this paper indicates that after a security event occurs in the S9 indicator, the reachable probability of each indicator in the experimental network shows an upward trend, and the overall security risk is rising, and at this time, the a posteriori reachable probability of S1, S4, and S6 is significantly higher than that of the other indicators, which is 0.85, 0.78, and 0.76, respectively, and it is very likely that there is a security risk in these three indicators. Comparing the a priori reachable probabilities of the indicator nodes given by the three methods, the a posteriori reachable probabilities of the indicator nodes of this paper’s method for S5, S7, and S8 are 0.46, 0.32, and 0.14, respectively, and there is no underestimation of the real security risk.
