Aiming at the problems of uneven material dispersion of traditional air classifier, NGZF classifier with downward air inlet fabric method is designed. In order to investigate the classification law of the classifier and the effect of secondary air on the flow field of the classifier, this paper utilizes ANSYS-Fluent19.2 software to simulate the internal flow field and conducts material experiments using the response surface method. The simulation results demonstrate that with the rise of the secondary air volume, the overall flow field distribution tends to be appropriate, and the washing flow field created by the secondary air does not interfere with the primary flow field. However, when the secondary air is too big, it will generate turbulence in the axial velocity in the guide cone zone. The discrete phase model (DPM) indicated that the secondary air would re-screen the misclassified fine particles. Response surface experiments with calcined petroleum coke showed that by setting the cut particle size at 50 μm and maximizing the classification accuracy, the ideal operating parameters were a rotor speed of 1037.4 rpm, a primary air volume of 429.6 m3•h-1, and a secondary air volume of 50.2 m3•h-1. This work not only provides a novel approach for the design and development of airflow classifiers but also gives theoretical direction for their implementation in industrial production.