Application of a radial basis function neural network to estimate pressure gradient in water–oil pipelines

Document Type


Publication details

Halali, MA, Azari, V, Arabloo, M, Mohammadi, AH & Bahadori, A 2015, 'Application of a radial basis function neural network to estimate pressure gradient in water–oil pipelines', Journal of Taiwan Institute of Chemical Engineers, vol. 58, pp. 189-202.

Published version available from:


Peer Reviewed



An accurate determination of the pressure gradient is required for efficient designing of oil and gas wells and pipe systems. Despite the recent improvements in accuracy of models and correlations developed for determining the pressure gradient, they are still incapable of estimating the pressure drop with desired accuracy. Therefore, a robust model is required to determine the pressure gradient precisely. Regarding high performance and great robustness of Artificial Neural Networks for solving science and engineering problems, this paper presents a Radial Basis Function Neural Network (RBF-NN) model to determine the pressure gradient. The model was developed over 994 experimental data sets which are covering a wide range of variables such as oil slip velocity, water slip velocity, pipe diameter, pipe roughness and oil viscosity.

The model estimation indicated an average relative deviation of 0.92%, an average absolute relative deviation of 8.25% and an average correlation factor of 0.99. A comparison between the proposed model and the most prominent models and correlations illustrated that the RBF-NN model exclusively out-performs other models and correlations and the estimated values are in great agreement with the experimental data. At last, a sensitivity analysis was applied to clarify the effect of input parameters in estimated results.