Modeling of the Blade Leading-Edge Pressure Drop of Centrifugal Impeller Based on Machine-Learning

Abstract

Blade leading-edge pressure drop is an important parameter that strongly influences the safe and stable operation of centrifugal pump. It is sensitive to blade leading-edge geometry. To understand the inlet flow state of centrifugal pump, it is necessary to evaluate the location and the minimum value of blade leading-edge pressure drop. However, the operation condition is complex. A large amount of experiments and numerical simulations should be conducted for a fully understanding of the local pressure field. With the development of artificial intelligence, machine learning can be used for accelerating the evaluations with potential laws of fluid mechanism. Therefore, neural network is used in this case to fit the solution space of condition, leading-edge geometry and local pressure drop. The influence of different factors can be well analyzed. The conditions which is not directly tested will be predicted by well-trained neural network. With the increasing of the leading-edge elliptic ratio of the blade inlet, the minimum pressure position of the blade moves from the impeller hub to the impeller shroud. This study provides a preliminary study and a good reference of the evaluation of blade inlet leading edge flow state in the future researches.