Multiphase Flow Monitoring in Oil Pipelines

Neural networks, and relateed statistical pattern recognition techniques, appear to be well suited to the solution of a wide range of monitoring and diagnos­tic problems. In many applications, it is difficult or impossible to perform first-principles modelling of the system under consideration. If, however, suffi­ciently large quantities of labelled training data<:an be made available, then a statistical approach becomes feasible.
In this paper we consider the non-invasive monitoring of oil flows along pipelines containing mixtures of oil, water and gas. This problem has arisen from use of multiphase pipelines to transport oil from offshore oil production platforms without the expense of offshore separation of the three phases. This in turn has led to the requirement for an accurate multiphase metering system for determining the oil fractions and flow rates for use in reservoir management and for custody transfer purposes.