Development of high-performance spray dryers that are more energy efficient and are able to produce high-quality milk powders is very important for the future of the dairy powder industry. Understanding and optimization of the exiting dryers are also of great value. Computational fluid dynamics is a powerful tool to simulate and help understanding the characteristics of spray drying and to introduce potentially improved designs. The present study has concentrated on the multiphase flow in an industrial-scale spray dryer using the CFD package FLUENT. A Eulerian-Lagrangian approach is used in the simulations. A new drying model, REA model, for milk particles has been implemented for the first time in a CFD application. The numerical results match well with the plant data. It is argued that the a??reflecting walla?? boundary condition produced more physically correct results for normal dryer operation than the a??escaping walla?? boundary condition. The second one is an oversimplification. The influences of the particle size and particle size distribution, residence time, kinetic energy, and maximum temperature have been analyzed. The initial droplet size range was set to be from 100 to 500 A?m with Rosin-Rammler distribution function. One significant result of this study is that rather dense particle clouds of medium-size particles (224-285 A?m) are found near the side walls. They are transported upwards along the side wall (conical wall surface) and then are dispersed. It is found the particles with initial diameters of between 225 and 270 A?m have the largest residence times.