This study investigates the effect of the particle injection position on particle classification performance of a cyclonic classifier. The Rankine-type vortex flow in the cyclonic classifier is divided into three characteristic regions from the wall to the center according to the tangential and axial velocity distribution. Three representative particle injection positions are selected within those regions. The particle flow and the particle residence time are investigated via the Reynolds stress and discrete phase models. The simulation result shows that particles injected in the near-wall region provide the highest separation efficiency but the lowest classification accuracy. The residence time of particles injected within the region of maximum tangential gas velocity is shorter, facilitating a rapid and efficient particle classification. The powder classification experiment result confirms the simulation result and the feasibility of this method to enhance the separation degree of particles in the vortex flow. (C) 2019 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.