In order to save the space for installation, a bent pipe is adopted for inlet of vertical inline pump. In this paper, to improve the performance of inlet pipe, a multi-objective optimization on the inlet pipe is proposed based on Genetic Algorithm (GA) and Artificial Neural Network (ANN) model. A 5th-order Bezier curve is applied to fit the mean line of the inlet pipe and 3rd-order Bezier curves are used for depicting the variation trend of shape of sections. As the outlet of inlet pipe is fixed, 11 design variables are utilized for optimization, and the three optimization objectives are efficiency, head and standard deviation of velocity at the outlet of inlet pipe. To get the surrogate model, 149 different models obtained from Latin hypercube sampling are solved with numerical simulation. The results showed the numerical simulation has a great agreement with the experiment. Artificial neural network can accurately fit the target functions and design variables. The deviation of efficiency, head and standard deviation of velocity between predicted value and actual value were 0.26%, 0.05m and −0.27m/s, respectively. After optimization, an improvement on flow condition and a decrease of standard deviation of velocity before impeller were obtained. The efficiency and head were improved by 1.16% and 0.2m, respectively.
- Fluids Engineering Division
Multi-Objective Optimization on Inlet Pipe of a Vertical Inline Pump Based on Genetic Algorithm and Artificial Neural Network
- Views Icon Views
- Share Icon Share
- Search Site
Gan, X, Pei, J, Yuan, S, Wang, W, & Tang, Y. "Multi-Objective Optimization on Inlet Pipe of a Vertical Inline Pump Based on Genetic Algorithm and Artificial Neural Network." Proceedings of the ASME 2018 5th Joint US-European Fluids Engineering Division Summer Meeting. Volume 1: Flow Manipulation and Active Control; Bio-Inspired Fluid Mechanics; Boundary Layer and High-Speed Flows; Fluids Engineering Education; Transport Phenomena in Energy Conversion and Mixing; Turbulent Flows; Vortex Dynamics; DNS/LES and Hybrid RANS/LES Methods; Fluid Structure Interaction; Fluid Dynamics of Wind Energy; Bubble, Droplet, and Aerosol Dynamics. Montreal, Quebec, Canada. July 15–20, 2018. V001T06A003. ASME. https://doi.org/10.1115/FEDSM2018-83053
Download citation file: