<p>To enhance the aerodynamic performance and efficiency of a centrifugal fan in a tea roller stirring machine, this study integrates bionics with computational fluid dynamics for a multi-objective optimization of the fan blades. The initial phase involves the adoption of a fish silhouette-influenced blade configuration protocol, entailing the extraction of characteristic ichthyic contours, and subsequently articulates through the class-shape function transformation (CST) for an integrated bionic blueprint. To improve the fan’s volumetric flow rate and overall efficiency in a scenario of zero static pressure, a carefully structured experimental array is created. This array includes parameters such as the impeller’s exterior and interior diameters, the radius at the arch point, the central angle of the blade, and the arch point’s central angle. The Latin hypercube response surface methodology is used to formulate this array, resulting in a diverse sample array of 30 sets. Following this, the moving least squares technique is harnessed to construct a response surface model, synergized with the global response surface method (GRSM) for the nuanced optimization of structural parameters. The outcomes indicate that the arch point radius predominantly affects the fan’s airflow volume, while the external diameter of the impeller essentially determines the whole machine’s efficiency. The optimized fan eventually achieves a notably increased airflow volume to 850.093&#xa0;m³/h, coupled with a 4.693% improvement in whole machine efficiency. This optimization significantly enhances the centrifugal fan’s capability to meet the requirements for tea leaf stirring processes.</p>