Modelling quality in spray drying

The objective of the research project described in this thesis is the development of a fundamental model that can be used for predicting product quality in spray drying processes. The approach to this is to model what particles experience in the drying chamber with respect to air temperature and humidity, in other words the particles' air temperature and humidity histories. These histories can be obtained by combining the particles' trajectodes with the air temperature I hurnidity pattem in the spray dryer. The model was based on a commercial computational fluid dynamics (CFD) package (CFX-F3D). The modeHing work and validation measurements were done fora co-current spray dryer (diameter 2.2 m; height 3.7 m; swirl angle 5°). The basic components of the model were stuclied in detail: the airflow, temperature and humidity pattems were modelled and measured; to study the partiele trajectories, the product residence time distribution was modelled and measured. The air temperature and humidity histories obtained with the model were used for the calculation of a product quality parameter, specifically the thennal degradation. The calculated degradations were compared with measurements. To prevent artefacts caused by an uneven in flow of air at the inlet, the swirl angle was set to zero for the CFD modelling and measurements of the airflow pattem. The modelled airflow pattem proved to consist of a fast flowing core and a slow circulation around that core. This agreed very well with experimental observations. The air inlet conditions for the CFD model were detennined by measuring with a hot-wire anemometer. This device was also used for velocity measurements in the drying chamber. There was good agreement between the measured and modelled airflow pattern. In the study of the temperature and humidity pattern, two different feeds were used: a feed of pure water, and one of an aqueous maltodextrin solution. The patterns were modelled by tracking particles through the flow domain and calculating the exchange of mass, energy, and momenturn along the trajectories. These exchange tenns were fed back to the airflow pattem calculation to obtain a two-way coupled solution. To calculate the exchange of mass and energy (evaporation), the drying kinetics of maltodextrin had to be incorporated in the CFD program. Because solving the diffusion equation would take too much computation time, a so-called short-cut metbod was used for this, which proved to be an accurate and efficient calculation method. Air temperature and humidity measurements were done to validate the modelling results. For this, a special device (a so-called micro-separator) was developed to prevent wet particles from depositing on the temperature and humidity probe and thus causing meaorement errors. The overall agreement between the modelled air temperature I humidity pattem and the measurements was good, but near the central axis there was room for improvement. The residence time distribution of the powder was modelled by tracking a large number of particles with turbulent partiele dispersion and recording the times of flight. By means of tracer analysis, the residence time distribution was measured. There was a large discrepancy: the measured residence time distribution had a long tail, caused by particles deposited on the conical part of the wall, foliowed by a slow sliding of the powder towards the product outlet. This sliding behaviour could not be predicted; nevertheless it was incorporated in the CFD model. This was done by assuming constant veloeities along the wall; the veloeities were obtained by fitting the measurements to the model. By combining the air temperature I humidity patterns with partiele trajectories, the particles' air temperature and humidity histories could be calculated. These were applied to the modeHing of a product quality parameter, specifically the thermal degradation of a-amylase. This was done by simultaneously solving the diffusion equation ( drying) and the degradation ra te equation; here the air temperature and humidity histories were used as time dependent boundary conditions. The calculations were done for three different air outlet temperatures. The results were compared with basic models and with measurements. There was little difference between the roodels and there was good agreement between roodels and measurements, provided that the sliding behaviour of the particles along the wall was included in the model. The thesis concludes with a discussion of the strengtbs and weaknesses of the CFD approach. When using CFD for quantitative quality predictions, a problem is the fact that it is very difficult to accurately measure the kinede properties of the product being dried. The strength of CFD lies therefore not in this type of calculations; the CFD approach can better be used for what-if studies and for gaining more insight in spray drying processes. As an example of a what-if study, the effects of increasing the swirl-angle from 5 to 15° on the residence time distribution is discussed.