Application Research Of Fuzzy Dynamic Bayesian Network In Drilling Risk

As the foundation of oilfield construction and development,the importance of drilling engineering is obvious.At the same time,because the economic cost of this project and there are a lot of uncertain factors,staffs need to keep all of rules in mind,strengthen prevention and manage,eliminate drilling risk and reduce the probability of economic loss to ensure that every employees will obey their work rules and promote the implementation of enterprise safety rules and regulations.Therefore,it is valuable to find and study approach to reduce drilling risk in drilling engineering.In this paper,the fuzzy dynamic bayesian network is applied to predict and evaluate the risk degree of drilling engineering.Fuzzy dynamic bayesian network is a theory combining fuzzy theory with dynamic bayesian network theory.In this paper,the introduction of fuzzy theory not only transforms traditional two-value bayesian networks into multivalued-value bayesian networks,but also relaxes the use of data.As a result,the establishment process and testing process of this model are as follows.The data used in this paper are from a historic data set of an oil field drilling project from January 1,2011 to December 30,2015.First,we identify risk factors based on historical risk data of drilling engineering,and construct their topological structure by these related risk factors.Second,according to the characteristics of the data set,we choose the K parabolic membership function,which has a membership function with a middle value of 1 and the change of the curves at both ends as the membership function,and explain the reason for the selection.At the same time,the data from January 1,2011 to November 30,2015 were used as the training set to calculate the network parameters,and the December 2015 data were left as the test set for the model prediction evaluation.The static bayesian network and dynamic bayesian network are constructed by topological structure and network parameter model based on risk factors.Then the data of the test set is divided into 10 test sets,which are predicted by the fuzzy dynamic bayesian network constructed respectively.We obtained a satisfaction prediction result and the comprehensive prediction accuracy is 72.857%.After that,we make an error analysis of the possible errors,and put forward some suggestions for the prevention of drilling engineering accidents,so as to strengthen the operation of the drilling operation and reduce the unsafe factors of the operation.