Application of a bayesian framework to decision making regarding maintenance and intervention of subsea water injection pumps

Abstract

As more and more oil fields in the North Sea are developed as subsea fields, the monitoring and maintenance availability is becoming difficult to perform on the equipment used for subsea boosting. Often maintenance is looked upon as a necessary evil (Arthur, 2005, Bevilacqua et al., 2003) rather than a way to optimize the production of the field. This paper is looking at the problem of when to pull a damaged Subsea Water Injection Pump and replace it. Usually the practice is to keep the pump running until it breaks down and the production has to be shut down while the pump is being replaced with a new one. However, this practice can result in very high costs due to shut down of the oil wells and bad weather during the intervention. The operator of the pump is receiving data from the subsea control system which carries information such as temperature, flow, head pressure, lube oil level, and vibration that describes the state of the pump. The data received is not easily interpreted due to scarcity and uncertainty of the data and these data cannot be complemented by a physical inspection of the equipment because it is located subsea. So a model that analyses the uncertainties and the events related to the pump and the intervention is needed. In this work the problem is formulated in a Bayesian framework and a decision analytic approach to analyze and determine when the pump should be pulled is being used.