Multi-objective shape optimization of offshore swirl-vane oil-water separators using genetic algorithm network coupled computational fluid dynamics

Compact and efficient oil-water separators are vital for offshore oil and gas extraction. Among these, swirl-vane separators play a crucial role, and their structural parameters significantly influence performance. In this study, a novel multi-objective optimization approach utilizing NSGA-II optimization algorithm and computational fluid dynamics is introduced to optimize the structural parameters of swirl-vanes specifically tailored for the polymer drive recovery process on offshore platforms. The Latin hypercube sampling method was initially employed for stratified sampling, facilitating the construction of a Kriging agent model trained on sample points. This agent model served as the objective function within the NSGA-II framework to establish constraints on structural parameters and derive a data set of Pareto-optimal solutions. Subsequent analysis focused on discussing the unique characteristics of the Pareto optimal points. Objective function values from these points were compared against those of the initial model. Results demonstrated notable improvements: optimal points A and B enhanced separation efficiency by 55.2% and 53.4% respectively, while reducing pressure drop by 36.1% and 38.7%. Optimal point C, despite a 37.9% reduction in separation efficiency, achieved a remarkable 64.5% decrease in pressure drop. The developed multi-objective method offers valuable insights for the parametric design of offshore swirl-vane separators.