Academic Research Of Drillstem Failure In Deep Well

In exploitation of deep oil and gas in the oilfield, it is necessary and urgent to solve the problemof lower drilling speed. When investigating the reasons for this phenomenon, we find that it ismainly because of the frequent drillstem failures in the oilfield. The support vector machinetechnology is a new technique for data mining and artificial intelligent information processing, ithas been developed based on the statistics learning theory, and it has a firm academic foundation.The SVM network is similar to the ANN;it can search optimization automatically and obtain theoptimal network structure by learning the limited sample data, meanwhile, it can overcome thedemerit of over learning of ANN. Clustering analysis is an important instrument for recognizingstatistical pattern;it can range a pattern into one kind of sorts or clustering classifications,therefore, patterns in the same clustering classification is more near to each other than differentones. The basic fundamental of clustering theory is on the view of birds of a feather flocktogether without former experiences;it can classify different kinds of samples into different sortsaccording to the size of the interval by using the mathematical techniques. Take view of the pointthat both of the two theories need to calculate the interval to a plane or a position, we propose anew technique to analyze the reasons for drillstem failure in deep well based on SVM and clusteranalysis, it can exert their predominances and obtain good results. Firstly, we introduce the maininfluent factors and basic types of drillstem failure in deep well, principle of SVM technologyand clustering analysis theory;secondly, we do some statistical work about the drillstem failureevents in Districts Xujiaweizi and Hailaer in Daqing oilfield, and we find out there are 80failures in 50 deep wells. By using the knowledge of statistical methods, we find out thestatistical rules of drillstem failures in deep well in Xujiaweizi and Hailaer districts;thirdly, byusing the SVM and clustering analysis we establish an analytic model for analyzing the reasonsof drillstem failure in deep well in oilfield, for instance we analyze the data come from DistrictsXujiaweizi and Hailaer and accomplish the rudimental reasons for drillstem failure in deep well.We develop a database software to record and restore the data, the software can input record,modify record, require record and output record at random according to well sites, years, districtsand failure types;we develop a drillstem failure analytical software of deep well. Both of thesoftwares provide convenience to the research. At last, we propose precautionary measures forDistricts Xujiaweizi and Hailaer based on the analytical results.