An Accident Prediction and Diagnosis Method Based on Bayesian Network

Against the limitations of some traditional accident prediction and diagnosis methods, a new method based on Bayesian network is proposed on the basis of the Fault Tree Analysis method. Using this method, accident prediction and diagnosis can become easily, quickly and more accurately. The method includes mainly two steps: the first step, to establish Bayesian network topology; the second step, to determine conditional probability table for each node. In addition, to update the marginal probability of the leaf nodes, in order to get a more objective evaluation, changing the past methods which mainly rely on statistics or expert experiences to give the probability value, it proposes an evaluation method using Bayesian network to combine different expert opinions effectively. Finally, taking the gas explosion accident as an example, using GeNIe Software to model and reason, taking analysis of the experimental data, we strive to prove the validity of the model.