A causality-guided and interpretable sweep–lean optimization framework of axial fan blades for aerodynamic enhancement

Axial fans are essential equipment in modern engineering applications, while variations in operating conditions inevitably increase energy consumption and adversely affect their aerodynamic performance. To enhance the aerodynamic performance of axial fans, an interpretable multi-objective optimization framework based on causal inference is proposed in this study. Eight sweep–lean control points of the blade are selected as design variables, and the heterogeneous treatment effects of different blade subgroups are analyzed using Causal Machine Learning (CML), through which the subgroups with accurate effect-oriented characteristics and the corresponding value directions of the sweep–lean control points are identified. Subsequently, based on the identified subgroup, a Kriging surrogate model is constructed in combination with the K-fold cross-validation method, and the interpretability of Partial Dependence Plots (PDP) is exploited to further reduce the search space of the design variables. Finally, global optimization is performed using the Non-dominated Sorting Genetic Algorithm II (NSGA-II). The results indicate that all eight design variables exhibit accurate causal effect guidance for blades with a medium-sweep and low-lean configuration, and the static pressure and total-to-static efficiency are increased by 1.79% and 2.23%, respectively. The axial momentum deficit of the airflow at the mid-to-high blade span is reduced, which effectively alleviates energy losses. The transitional flow between the wake and the mainstream is significantly improved, thereby suppressing the lateral expansion of the wake.