The Research And Implementation Of Prediction Of Drilling Failures Based On Data Mining

In drilling platform’s production process,drilling equipment and drilling environment from down hole will result in drilling failure.Drilling failures can be divided into different types according to the severity,which may cause drilling equipment failure or even drilling accident.Serious drilling failures may result in closure wells and other consequences,so it is necessary to monitor the drilling failures.At present,the existing way to prevent drilling failures is manual monitoring,which is observing the fluctuation of some specific parameters.However,the method of parameter monitoring is overly dependent on the fluctuation of single or several parameters,and the occurrence of each type of drilling failure has a complex relationship with multiple parameters.In view of this situation and based on the actual drilling failures’ data,this paper studies the drilling fault identification and warning method based on the classification algorithm.The main work is as follows.(1)According to data characteristics and combined with the application of Two-Step clustering algorithm and preferred parameters,the paper completed the cluster analysis.Then according to the real record of the label,the paper done the work of replacing labels.The trend of drilling failures in the source data can be revealed in the data labels by using this method.Finally,the source data is complemented to lay a solid foundation for subsequent model training.(2)After cleaning data and selecting parameters for data sets,the paper used C4.5 decision tree algorithm to obtain the early warning model of drilling failures.Finally,this paper recognized the accuracy of the model and the effect of early warning and compare with existing methods.Through the practical study of this article,this thesis proved the validity of Two-Step clustering algorithm and C4.5 decision tree,and the recognition accuracy of the model reaches more than 90%.