Generally, modelling approaches can be divided into two major groups: first principle models are involved wherever possible, relying on heat, energy and population balance equations. For the most complex processes where the application of first principle models would be too difficult or nearly impossible, multivariable statistical and machine learning methods are employed. In certain cases, hybrid mechanistic-statistical models are also developed.
Our reaction modeling expertizes include homogeneous reactions occurring in batch and continuous (tank and tubular reactors), which was demonstrated by the identification of the reaction network describing the synthesis of acetyl salicylic acid. Intensive crystallization modeling work is performed using mono and multidimensional population balance models, covering the crystallization of isotropic and high aspect ratio crystals, solution mediated polymorphic crystallization and transformation and preferential crystallization of enantiomers. Pharmaceutical filtration and drying processes were also described with semi-empirical first principle models recently.
The application of artificial neural networks as black-box models were successfully demonstrated in the case of residence time distribution modeling of a pharmaceutical twin-screw wet granulator.