An Improved Data-Driven Model for the Prediction of Minimum Transport Condition for Sand Transport in Multiphase Flow Systems

The correct prediction of minimum transport condition (MTC) is of great importance to the oil and gas industry. The sand deposition is an associated problem of multiphase transportation of oil, gas and or solid. The purpose of this work is to investigate the predictive capability of three different data-driven approaches: Artificial neural networks (ANN), Adaptive neuro-fuzzy inference system or adaptive network-based fuzzy inference system (ANFIS) and response surface methodologies (RSM). The models were developed using182 experimental data points with input parameters such as liquid superficial velocity, pipe diameter, particle size, pipe inclination and the output parameter predicted is the minimum transport condition (velocity) for sand particles. The developed models were compared with existing models. The results showed that the three methods performed creditably well in the prediction of MTC with ANFIS having the highest predictive capability with an R 2 value of 0.99997 and an average error value of 0.00035836 compared with ANN and RSM having R 2 value of 0.9998 and 0.9973 respectively. The three data-driven techniques investigated in this study also outperformed published correlations for the prediction of MTC. The findings from this research can be invaluable for the effective and robust management of sand transport in multiphase flow systems. Keywords — Artificial Intelligence, Fuzzy Inference System, Model, Minimum Transport Condition, Optimization methods, Response Surface Methodology