Model Ranking Methodology for Accurate Pick-up Velocity Predictions to Initiate the Motion of Sand Particles At Low Sand Particle Concentration In Horizontal Single Phase Fluid Flow

The incipient motion of sand particles from the bottom of pipelines has been a topic of interest due to the desire to prevent the accumulation of sand particles in pipes. In order to initiate the movement of the sand particles, the magnitude of the fluid velocity must be equal to at least that of the pick-up velocity of the sand particle. There are multiple sand transport models that predict the pick-up velocity, but for the same operating conditions, the velocity estimations of these models might range over four orders of magnitude. Therefore, given a set of operating conditions, it is necessary to determine the models that best predict the pick-up velocity for those conditions.

This need is addressed with the introduction of a novel model optimization and ranking methodology that combines data clustering approaches with statistical analysis. The data clustering component collects the representative data points that lie within the range of the operating conditions, and the models are optimized using those data points to ensure that they perform optimally within the range of the operating conditions. Then, those models are ranked to determine the ones that make the most accurate velocity predictions within the range of the operating conditions. Lastly, the uncertainty of the velocity predictions of the models is quantified to determine the upper bound of the velocity predictions, thus providing more conservative results. The methodology is tested using field operating conditions, and the results yield consistent pick-up velocity predictions when compared to the observed ones.