Pore-pressure estimation is an important part of oil-well drilling, since encountering unexpected highly pressured fluid can be costly and dangerous. Despite this, standard estimation methods do not account for the many sources of uncertainty, or for the multivariate nature of the system. A collection of simplistic deterministic functions is used, and input parameters are often `fudged' to obtain the desired results.
I will present our Bayesian network model of the pore pressure system, which addresses these issues. We model the relationships between quantities in the pore pressure system, such as pressures, porosity, rock type and log data, using conditional probability distributions derived from geophysical relationships to capture our uncertainty about these quantities and the relationships between them. Our model provides a coherent statistical framework for modelling the pore pressure system with uncertainty.