Failure prediction in open-hole wireline logging of oil and gas drilling operation using support vector machine.

Abstract

The failure of open-hole wireline logging leads to an unexpected cost and time that add to drilling operation. The research proposes methods to predict the failure of an open hole wireline logging prior to run the log on actual situation. Three machine learning techniques are used to classify the result of the open-hole wireline logging from drilling process into two classes, a success class and a failure class which represents a well that might have abnormal conditions which can causes the tool stuck during logging. The success class is the normal well that can run logging to target depth without tool sit down or stuck. Support Vector Machine, Naive Bayes and Decision Tree are chosen as proposed machine learning techniques for this classification.