Separation efficiency prediction of non-Newtonian oil-water swirl-vane separators in offshore platform based on GA-BP neural network

Abstract

Oil and gas processing plays a crucial role in offshore oil production platforms. Swiftly predicting the effects of fluid parameters on the efficiency of oil-water separation is essential for enhancing the performance of swirl-vane separators. To investigate the behavior of oil-water two-phase flow, a Eulerian multiphase flow model was employed in this study coupled with the power-law non-Newtonian fluid model. Input variables including the oil volume fraction, oil density, inlet velocity, oil droplet particle size, oil droplet viscosity coefficient, and oil rheological index were selected with separation efficiency as the output variable. Based on the genetic algorithm, the backpropagation (BP) neural network was optimized to predict the effects of fluid parameters on the separation efficiency of a swirl-vane separator. The results demonstrated a remarkable improvement in prediction accuracy, exceeding 50 % at its maximum. Notably, oil density and volume fraction exhibited significant impacts on the swirl-vane separator. A multiple linear regression analysis of experimental and simulation results revealed that the volume fraction, inlet velocity, oil droplet particle size, and oil phase rheological index are positively correlated with separation efficiency. Conversely, the oil density and oil viscosity coefficient exhibited a negative correlation with separation efficiency.