Controlled agglomeration during spray drying offers several advantages for both powder manufacturers and consumers, and thus it is commonly implemented by industry. The implementation, however is largely based on experience, given the scarcity of comprehensive prediction tools. A resource-efficient approach to numerically treat agglomerates and yet provide an indication of their structures is desired to perform realistic simulations without the need for high-performance computing. In this work, a new numerical model for the treatment of coalescence and agglomeration was implemented and evaluated at two distinct scales with significantly different particle number densities within a Eulerian-Lagrangian CFD framework. The model could accurately predict the trends in the final particle size distributions and distinguish realistic agglomerate structures occurring under different conditions. Challenges were encountered as a result of how the underlying collision detection routine handles high particle number density. Several strategies are proposed to overcome these challenges. This work constitutes significant progress towards achieving an efficient prediction tool to estimate final powder properties and will prove useful in performing large-scale simulations to design and control agglomeration. (C) 2020 Elsevier B.V. All rights reserved.