8 th International Conference on Multiphase Flow ICMF 2013, Jeju, Korea, May 26 - 31, 2013 PERFORMANCE ASSESSMENT STUDY OF MULTIPHASE FLOW MEASUREMENT WITH CORIOLIS FLOWMETER USING PATTERN RECOGNITION TECHNIQUE Eni Oko a , Hoi Yeung a * and Tesi Arubi b a Process Systems Engineering Group, School of Engineering, Cranfield University MK43 0AL Bedfordshire, UK b BP International, Sunbury, UK *Corresponding author. T.: +44 (0) 1234 758266. E.: h.yeung@cranfield.ac.uk (H.Yeung). KEY WORDS: Artificial Neural Networks, Differential Pressure Sensor, Stochastic Features, Gas Void Fraction Abstract This paper reports an assessment of Coriolis mass flowmeter for multiphase flow measurement using pattern recognition (PR) techniques. Coriolis mass flowmeters do not offer acceptable performance for multiphase flow measurement where the gas void fraction (GVF) exceeds 5-15%. However, if the relationship between the measurement errors arising from increasing GVF could be investigated, appropriate models can be developed and used to improve its performance at higher GVF (>15%). Such models have been developed in this study through PR involving artificial neural networks (ANN). This was done using stochastic features obtained from the amplitude and frequency domain of differential pressure (DP) and Coriolis signals from a V-cone meter and a Coriolis mass flowmeter respectively. These signals are acquired over different multiphase flow conditions at the Cranfield University three-phase flow test facility. The prospect of obtaining more accurate PR model by data fusion involving a combination of DP and Coriolis features is also investigated. The performance of the PR model developed using Coriolis features showed that within 5% target error, 94%, 92% and 90% of the test points covered were predicted accurately for gas volumetric flow rate, liquid mass flow rate and water cut. The PR model obtained using DP features predicted 87%, 100%, and 87% of the test points for gas volumetric flow rate, liquid mass flow rate and water cut respectively within the same error margin of 5%. PR model based on the combined features proved to be the most accurate predicting 99%, 100% and 100% respectively of the test points for gas volumetric flow rate, liquid mass flow rate and water cut within 5% error margin. In this study, we demonstrated the potentials of Coriolis mass flowmeter for multiphase flow measurement using PR modelling and the possibility of improving such measurement using data fusion. This is an initial study however and requires more detailed study in order to obtain more conclusive result. 1. Introduction Multiphase flow metering in oil and gas production facilities is necessary for reservoir management, production monitoring and optimization, well performance monitoring and intervention among others. Metering data facilitates better planning and policies which translates to cost savings. Traditional multiphase flow metering involving the use of test separators has huge cost and space requirements (Arora, 2009). Reports by Jamieson (1998) show that the costs of using test separators are significant compared to multiphase flowmeters (MPFM). Also, as oil and gas exploration enter into more difficult offshore terrains and the need for economic