Monitoring and control of continuous multi-compartment fluidized bed spray granulation: Real-time predictions and flow-topology based process actuation
Even though the fluidized bed spray granulation has been applied for more than 60 years, it is still a challenge to control the product particle size. The aim of our project is the development of a real-time monitoring and prediction framework, which enables the control of the product particle size distribution in continuous multi-compartment fluidized bed granulation processes. This objective will be achieved by three innovative pillars: novel switching-topology for fluidized bed operation, real-time simulations allowing for online process control and on online particle size measurement setup. Central part of the experimental investigations will be a two-stage lab-scale fluidized bed equipped with an underflow weir. We propose to use the phenomenon of segregation to influence which particles enter the spray zone and thus increase in particle size. By switching from homogeneous aeration to heterogeneous aeration, the topology of the particle circulation pattern will be changed. In homogeneous aeration mode, only larger particles are handed over to the consecutive stage by an underflow weir while smaller particles remain in the top-spray zone. The injection of a higher velocity gas stream (heterogeneous aeration) will cause circulation within the fluidized bed and thus homogenize the particle size distribution. Large particles will grow further just the same as smaller particles. As a novel measurement concept, we intend to assess the particle-size-distribution while particles pass the underflow gate between individual stages. To this end, the underflow weir gate is equipped with a mirror and an imaging system. By taking periodic snap-shots we will receive images of a set of particles on their flight through the gate. Online digital image analysis will provide instantaneous PSDs. The control of the spray granulation process will be performed by real-time simulations of fluidized bed dynamics by the recurrence CFD (rCFD) approach. rCFD uses recurrence plots of short-time CFD-DEM simulations (recurrence database) to time-extrapolate the fast dynamics based on the assumption that similar states will have a similar near-time evolution. Using this method, the internal state of the system can be simulated faster than real-time, which is not possible with pure CFD-DEM simulations. In this project, this radically new methodology will be further developed for poly-disperse systems. For this purpose we have to track and modify local particle size distributions. The fluidized bed control system will be integrated with the rCFD solver to synchronize the system state with choice of the database. This includes communication of outlet air temperature and humidity, as well as particle size distribution determined from the gate measurement setup and an interface for the control of the switching mechanism. Thus, the predictions from the rCFD solver will be used to choose the correct flow topology to yield the desired product particle size distribution.