Hydro energy is a kind of typical renewable energy, which can be converted by hydraulic machinery. However, tip leakage vortex (TLV) has a significant negative influence on the flow pattern and energy performance of hydraulic machinery. In this paper, a bending shrinkage groove (BSG) is proposed to suppress the TLV and improve the energy performance of a hydrofoil first, and then a parametric optimization design method is briefly introduced and applied to determine the optimal configuration of the groove. The main geometric parameters of the groove are selected as optimized variables and three different groove configurations are selected from the optimization result. Finally, the performance improvement of the hydrofoil with groove, the sensitivity analysis of the optimization variables, and the groove impacts on the TLV and flow patterns are investigated. The results demonstrate that the preferred groove reduces the non-dimensional Q criterion vortex isosurfaces area (Q_area = 2 × 10⁷) by 5.13% and increases the lift drag ratio by 17.02%, comparing to the origin hydrofoil. Groove depth d and groove width w are proved to have more significant impacts on the hydrofoil energy performance. The TLV and flow patterns are greatly affected by the different BSG configurations, and the wider BSG contributed to reducing the area of TLV, at the cost of energy performance deterioration.

The traditional design method, design-experiment-design, is time consuming and difficult. In the present paper, an efficiency design system for pump-turbine runners, the multiobjective optimization design system, which includes 3D inverse design, Computational Fluid Dynamics (CFD), Design Of Experiment (DOE), Radial Basis Function (RBF) and Multi Objective Genetic Algorithm (MOGA), is presented and then applied to design a scaled ultrahigh-head pump turbine runner. First, the control parameters, blade loading and the splitter work ratio, and the runner efficiencies under pump and turbine mode were selected as input variables and optimization objectives respectively. Then the DOE was applied to select the sample runners, and the CFD was used to predict the efficiencies of the sample runners under different modes. The relationship between input variables and objectives was built by RBF based on CFD results. Finally, the MOGA was applied to generate and search the good overall performance runners according to the relationship built by RBF. The results indicate that: the pump mode efficiency of preferred runner is increases by 0.34% and the turbine mode efficiency by 2.07% compared to the initial runner. With the improvement of efficiency, the pressure and streamline distribution are improved obviously.

Deep learning models provide a novel research perspective for hydraulic machinery and fluid dynamics mechanism research. Traditional computational fluid dynamics requires a lot of computational resource and calculation time, while deep learning models can effectively solve this problem. In this paper, a deep learning model is proposed for the rapid flow field analysis of a two-dimensional cylindrical bypass flow, and the errors of the prediction results are analyzed, so as to verify the feasibility of deep learning for accelerating the numerical simulation process. On this basis, the influence of different network structures on the prediction performance of the deep learning model is explored, and the optimal structural parameters of the neural network are found, indicating that it will achieve real-time prediction of the flow field performance, and save considerable computational resource and calculation time. The research in this paper is of great significance for the application about the rapid analysis of hydraulic machinery fluid dynamics based on deep learning models.