Integrated Lost Circulation Prediction Using Support Vector Machine in Iranian Oil Field

Lost circulation and stuck pipe are the most common drilling problems which always have been challenging. Mechanical pipe sticking is likely to occur after complete loss. Lost circulation is likely to take place in throughout drilling operation and even while primary cementing, because of direct contact of the drilling fluid with formation along with severe pressure pulses due to pipe movement or onset of circulation after connection which sometimes goes over hundreds of psi. Thus, having accurate information about returned fluid and recording mud loss rate can be great help to prevent drilling problems from taking place. Although recent solution to deal with lost circulation is directed to under balanced drilling technique, this is not applicable in some countries due to lack of technology or requires huge expenses. Thus, prediction of loss severity can bring the opportunity of decision making true for adjusting drilling fluid content and operational parameters. Several factors while drilling will govern how severe mud loss would occur. These actually make analytical modeling of lost circulation to somehow complicated. Hereby, employing Support Vector Machine (SVM) can be a leeway with proven capability and accuracy. In this research, operational parameters in one of Middle Eastern oilfields are used for prediction of the mud loss severity along different sectors of this oilfield. Performed cross validations and comparison with artificial neural network show good compatibility with what happened in reality.

Integrated Lost Circulation Prediction Using Support Vector Machine in Iranian Oil Field Keywords:

Virtual Intelligence , Support Vector Machine , Lost Circulation , Middle East