The present study investigates the design of a very low head axial ow turbine using surrogate-based optimization. The design variables were blade angles between guide vanes and runner blades with the objective function of turbine eciency. A Latin hypercube sampling method was initially used to design the experiment with thirty sampling points, and a large eddy simulation was modeled to analyze the ow for all sampling points. The correlation between the design variables and turbine eciency was then evaluated using surrogate models while the optimal design variables were identi ed. In addition, several optimizers were employed to tackle the proposed problem and evaluate their performance. The optimal design of blade angles.