Predicting the pumping characteristics of multiple parallel tube air-lift pumps

Abstract

Air-lift pumps have begun to receive a high degree of attention due to the absence of mechanical components and the potential for their use in renewable energy applications. One of the principal challenges of the air-lift pump is increasing the volume of fluid it can pump, as such it may be possible to utilise multiple parallel tubes. In such an arrangement it is necessary to have the two phases distributed to multiple tubes from a common source. However, from an analytical perspective this leads to multiple steady state solutions and hence accurately predicting the pumping characteristics of an air-lift pump becomes extremely complex.

To circumvent the analytical challenges associated with dividing a multiphase flow amongst multiple parallel tubes this work utilised an artificial neural network (ANN) (a class of artificial intelligence) to the prediction of the pumping characteristics of an air-lift pump with multiple parallel lift tubes. The results show that the neural network model provides an extremely accurate prediction of the pumping characteristics of multiple tube air-lift pumps within the training bounds. Moreover, the ANN provides insights into the pumping characteristics of multiple tube air-lift pumps outside these bounds that would be extremely difficult to achieve by analytical means