Online microscopy has received much attention in the field of crystal shape control over recent years, since commonly used measurement techniques cannot provide enough information for this purpose. In this work, we present an estimation scheme that serves to reconstruct the 3D crystal shape from the measured 2D crystal projection. The boundary curves of the crystal projections are parametrized by Fourier descriptors, which are subsequently compared with a precomputed database. The procedure is evaluated in comparison with various effects that might impair the estimation. A good agreement between the true and estimated values is found in all cases. The presented methods are applied to batch cooling crystallizations of potassium dihydrogen phosphate (KDP) dissolved in water. As a result, the face specific growth rates are determined as a function of supersaturation. To validate the performance of this scheme, the calculated face specific growth rates are used in a model-based prediction of the supersaturation profiles, which agrees well with the experimental data.

The aggregation of crystals within a flow tube was observed based on data extracted from images of the bypassing population. The experiments were conducted under different conditions, namely the flow rate and the particle concentration have been varied simultaneously and two different solvents were used in which the aggregation extent was found to be different under otherwise constant conditions. The analysis of images of bypassing crystals allows for the acquisition of rich datasets both in terms of the variety of shape descriptors and number of particles. This amount of data enables the determination of at least bivariate number distributions of high accuracy with simple histograms. The interpretation of the data is further improved with kernel histograms with which also higher-dimensional volume distributions can be obtained in good quality based on relatively few data points. Indeed, the isosurfaces of 3D distributions turned out to be helpful for inspection of the acquired data.

No abstract

No abstract

No abstract

No abstract

No abstract

Crystallization processes are characterized by a close interaction between particle formation and fluid flow. A detailed physical description of these processes leads to complicated high-order models whose numerical solution is challenging and expensive. For advanced process control and other model-based online applications, reduced-order models are required. In this work, a reduced model for a urea crystallizer is developed using the method of moments for the internal coordinate and proper orthogonal decomposition for the external coordinate. Simulations are carried out to compare the reduced model with the detailed reference model.