Optimization of U-shaped flow channel by RBFNN and NSGA-II

Abstract

Mobile machinery is in desperate need of weight reduction. To achieve this goal, their hydraulic reservoirs are developing in the direction of miniaturization and lightweight. However, this development of the hydraulic reservoir can lead to excessive air content in the hydraulic system and hard pump suction. At present, how to solve this conflict has not been studied and how the hydraulic reservoir structure influences the deaeration and inside pressure drop is not clear. In this paper, a U-shaped flow channel in a minimized hydraulic reservoir was numerically optimized and analysed for efficient deaeration of hydraulic systems and minimal inside pressure drop. Based on the multiphase flow simulation database, radial basis function neural networks and an NSGA-II were combined for the flow channel optimization. Results showed that two optimized structures raised the degassing rate by 8.8% and 18.1% and reduced the pressure drop rate by 23.7% and 13.5%, respectively. The flow field analysis and bubble motions were investigated among the initial and optimized flow channel structures. On the baffle's left side, bubbles were degassed by the buoyance force. On the baffle's right side, the wall confinement bubbly jet transformed bubbles to the free surface and degassed, but this jet formed by the deep baffle would increase the pressure drop, which agrees with the parameter sensitivity result. Additionally, vortex structures caused by the jet flow entrained bubbles and impeded deaeration. Consequently, the strong vortex was suppressed in optimum models to increase degassing capacity.