Optimal model-based experimental design in batch crystallization

The model-based experimental design of batch crystallizers is investigated. A dynamic programming formulation minimizes the volume of a confidence hyperellipsoid for the estimated nucleation and growth parameters over the supersaturation profile and the seed characteristics, namely, the crystal mass, mean size, and width of the seed distribution. It is shown that the accuracy of the parameter estimates can vary by several orders of magnitude depending on the seed characteristics, and that highly accurate estimation of nucleation and growth parameters can be obtained with as few as four batch experiments.