MaxEnt 94

MaxEnt94 Abstract

Bayesian event detection

Tim Jervis

Cambridge University Engineering Department and Schlumberger Cambridge Research

Drilling for oil is a hazardous enterprise. If gas leaks into the well during drilling, and is unnoticed, a chain of events takes place which can lead to the drilling rig exploding and a subsequent loss of life. Early detection of the influx means the well can be controlled safely. I shall describe current approaches to the early detection problem used in the oil industry, and a simple Bayesian approach that out-performs these algorithms on real data.
Cambridge University Engineering Department
Trumpington St.
Cambridge CB2 1PZ