Improved crystallization processes benefits industry
Crystallization is one of the most important production and separation processes in the chemical and pharmaceutical industry. However, due to unpredictable crystallisation mechanisms, industrial crystallisers are seldom operated under automatic control schemes. This frequently leads to non-reproducibility, unacceptable product qualities and excess energy consumption. A significant increase in both the efficiency and effectiveness of crystallisation processes was the main objective of the SINC-PRO project, Self learning model for INtelligent predictive Control system for crystallisation PROcesses. The improvement of effectiveness refers to desired product quality whereas the increase of efficiency would result in a reduction in costs. Model-based advanced control algorithms such as model predictive control and neuro-fuzzy control were implemented in order to develop the techniques for the on-line control of industrial crystallisers. Both these control algorithms can be continuously updated by applying a self-learning technique. A flexible process-modelling tool, covering a wide range of crystallisation processes, will provide the tailored models required for model-based control. Furthermore the most suitable actuators and parameter observation tools are selected in order to manipulate the process variables and provide an integrated control system. The control system was tested on laboratory scale with different combinations of measurement systems and actuators, giving very encouraging results. Moreover, the field tests at the sites of the industrial partners proved that model predictive control can be used in order to control crystallization according to a predefined process trajectory. This evidence renders this technique an available option for the industry.
