Modeling batch crystallization processes: Assumption verification and improvement of the parameter estimation quality through empirical experiment design

Abstract

In this work, experimental data of different batches was used for estimation of the kinetic parameters for the secondary nucleation framework of Gahn and Mersmann [Gahn, C. and Mersmann, A., 1999. Brittle fracture in crystallization processes. Chem. Eng. Sci. 54, 1273–1292]. An empirical experiment design procedure was used to design an informative batch experiment through optimization of the seed quality, size and mass and process conditions at seeding. The parameters estimated using the data of the designed experiment showed smaller magnitudes of the confidence ellipsoids and standard deviations as compared to those obtained by using the data of conventional (un)seeded batch experiments. It was shown that the designed experiment allowed reducing uncertainty in the initial conditions, namely, the mass and crystal size distribution of the initial population of crystals and the initial supersaturation. It was also demonstrated that the main reason for the model/process mismatch was the origin of nuclei. Dynamic experimental data could be described better if the state of the crystals forming the crystallization system corresponded to the assumptions of the used kinetic model. Differences in the crystal surface properties, shape, and strain content could be responsible for a divergent nucleation and growth behavior in batches that were initiated either by primary nucleation, seeding with small ground seeds or seeding with coarse crystals from the product of the previous batch.