TY - JOUR
T1 - Two-phase flow pressure drop modelling in horizontal pipes with different diameters
AU - Faraji, Foad
AU - Santim, Christiano
AU - Chong, Perk Lin
AU - Hamad, Faik
PY - 2022/7/1
Y1 - 2022/7/1
N2 - The two-phase frictional pressure drop has a dominant effect in many industrial applications associated with the multiphase flow. This study investigated the accuracy of several available methods for predicting two-phase frictional pressure drop of different pipe diameters using 4124 experimental data points. It is observed that the performance of the existing methods is poor in a wide range of operating conditions. Then, several Artificial Neural Network models were proposed, including six multilayer perceptron (MLP) and one Radial Basis Function (RBF) using the same data sets. The weights and biases of the ANNs were optimized using Levenberg-Marquardt (LM), Bayesian Regularization (BR), Scaled Conjugate Gradient (SCG), Resilient Backpropagation (RB), Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). Statistical error analysis indicates that neural network incorporated with the Genetic Algorithm (MLP-GA) predicts the entire data set with a Root Mean Square Error of 0.525 and an Average Absolute Relative Error percentage of 6.722. Finally, the sensitivity analysis was carried out, indicating that the mass flux (G) has the highest direct impact on the two-phase frictional pressure drop.
AB - The two-phase frictional pressure drop has a dominant effect in many industrial applications associated with the multiphase flow. This study investigated the accuracy of several available methods for predicting two-phase frictional pressure drop of different pipe diameters using 4124 experimental data points. It is observed that the performance of the existing methods is poor in a wide range of operating conditions. Then, several Artificial Neural Network models were proposed, including six multilayer perceptron (MLP) and one Radial Basis Function (RBF) using the same data sets. The weights and biases of the ANNs were optimized using Levenberg-Marquardt (LM), Bayesian Regularization (BR), Scaled Conjugate Gradient (SCG), Resilient Backpropagation (RB), Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). Statistical error analysis indicates that neural network incorporated with the Genetic Algorithm (MLP-GA) predicts the entire data set with a Root Mean Square Error of 0.525 and an Average Absolute Relative Error percentage of 6.722. Finally, the sensitivity analysis was carried out, indicating that the mass flux (G) has the highest direct impact on the two-phase frictional pressure drop.
UR - https://www.sciencedirect.com/science/article/pii/S0029549322002175
U2 - 10.1016/j.nucengdes.2022.111863
DO - 10.1016/j.nucengdes.2022.111863
M3 - Article
SN - 0029-5493
VL - 395
JO - Nuclear Engineering and Design
JF - Nuclear Engineering and Design
M1 - 111863
ER -