Downhole failures revealed through ontology engineering

Abstract

Frequency of failures occurring during offshore oil-well drilling do not diminish over time, probably due to increased operational complexity. The volume and frequency of information generated during the drilling process are high. Every few seconds around 30 drilling parameters are recorded and stored. There will always be a need for new, smart solutions to drilling challenges. Present approaches to drilling challenges apply various types of knowledge generated by the operation; wellbore geometry, fluid info, lithology of the sediments, time-based drilling parameters, drilling equipment data, etc. In our research said knowledge is generalized into general concepts, and structured to form a knowledge model of the drilling process. This model is referred to as a drilling ontology, and we report how methods of knowledge modeling and ontology engineering have been used in developing the model, and subsequently how the model has been applied to predict downhole failures during drilling.

The knowledge model and the drilling data are combined in the following manner: Data agents are surveilling drilling data. Deviatoric behavior of the drilling parameters are being detected and formed into symptoms. Symptoms trigger other concepts embedded in the ontology by means of linked cause-effect relationships. The end concept of the relationship path will always be one or several failure states.

Tests show that reasoning within the ontology produces the highest probability of the failure identical with the real failure. The causes behind the failure can be retrieved from the ontology and applied in a useful manner in combatting the failure. The testing process also shows that this program is a potential supplement to warning against threatening failures before they occur.