Three-dimensional multi-objective aerodynamic optimization of blade metal angle matching in multi-stage axial fan

The blade metal angles in multi-stage axial fans are key factors determining blade additional work and efficiency. Conventional blade matching designs typically rely on one-dimensional and two-dimensional methods, which are highly dependent on empirical models. Therefore, it is of considerable importance to conduct multi-blade-row metal angle matching optimization based on three-dimensional simulations. This paper adopts a surrogate-assisted multi-objective optimization framework, taking the last three blade rows of a three-stage axial fan as the optimization target. A three-dimensional blade parameterization tool is coupled with a multi-population differential evolution algorithm (MPDE) to reconstruct the original geometry with high fidelity by minimizing the Chamfer distance, and the inlet and outlet metal angles at three critical spanwise sections are selected as design variables. High-fidelity Reynolds-averaged Navier-Stokes (RANS) simulations are performed at 100% and 80% of the design rotational speed to generate the training dataset. An artificial neural network surrogate model is trained using five-fold cross-validation for hyperparameter tuning, and the non-dominated sorting genetic algorithm II (NSGA-II) is subsequently employed to search for the Pareto-optimal front. The optimization results show that the overall isentropic efficiency of the three-stage fan is improved by 0.61% at the design rotational speed and 0.81% at 80% design rotational speed, respectively. These results validate the feasibility and effectiveness of multi-objective aerodynamic optimization based on three-dimensional simulations of multistage fan blades.