Dosta, MaksymMaksymDostaTsz Tung, ChanChanTsz Tung2022-02-212022-02-212022-01Powder Technology 398: 117156 (2022-01)http://hdl.handle.net/11420/11733To improve predictivity of macroscale flowsheet models and to establish a link between process conditions, material microstructure and product properties, a data-driven strategy is proposed and applied for continuous particle formulation process. A discrete element method and mesh-free bonded-particle model are used to analyze mechanical behavior of multicomponent agglomerates at uni-axial compression tests. The DEM calculations are performed for varied input parameters to create a database containing information about fracture behavior of agglomerates. The final database is used to build an artificial neural network (ANN) and to link structure-property relationships: from known properties of single components and known microstructure to predict macro-mechanical agglomerate properties. Afterward, the formulated ANN is coupled to the population balance model (PBM) to perform modeling of continuous process where the transient change of particle size distribution in the plant is described. The results demonstrate that the proposed strategy can be efficiently applied to link process-property relationships.0032-59102022Artificial Neural Network (ANN)Data-driven simulationDiscrete Element Method (DEM)Multicomponent agglomeratesPopulation Balance Model (PBM)Journal Article10.1016/j.powtec.2022.117156Other