Application of fuzzy dynamic bayesian network in drilling risk(in Chinese)
This thesis discusses some risks of drilling and management in oilfield, as well as a way for preventing oilfield accidents.
A oilfield records its accident data from 2011 to 2014.
And those data are classed into different factors due to the cause.
The data in different factors have different fluctuation.
For example, the maximum value or minimum value in different may be different.
This means that it is hard for the inexperienced manager to distinguish the risk rank.
In order to discussing the risk rank, the membership function(a fuzzy theory)was used to evaluate it in different factors.
The risk rank is simply classed into three categories in this thesis: high risk, middle risk and low risk, which will decide the data risk rank.
On other hand, bayesian network is feasible and proper tool to predict data trend, although it usually based on the probability.
The risk factors based on their similar cause are structured as a hierarchical network.
The structure was built in GeNie before the data were inputted.
Bayesian network are a special kind of probility model because it usually provides a positive feedback to the 'cause'.
The prediction accuracy is over 73% which is considered feasible.
