Intelligent control of crystallisation processes
The majority of current industrial crystallisation processes are rarely operated automatically since effective and efficient control of product size distribution are hard to achieve. Hence, the process is not reproducible, is excessively energy consuming and results in potentially unacceptable quality of products. Urged by this the SINC_PRO project implemented a self-learning model for an intelligent predictive control system for crystallisation processes. This innovation includes a flexible process modelling tool, an observer/feedback system integrated with on-line measurement techniques and a toolbox including two types of control - model predictive and neural network control. Opting for the most appropriate modelling approach, researchers concentrated on developing a software package with increased extendibility that can be employed under various working environments. For this reason, the final architecture features a data server structure that couples with various software tools, each offering rigorous simulation, observation and control. Another important requirement was to enable proper process measurements and actuators for the control of crystallisers. Key process measurement properties include saturation, crystal mass and crystal size distribution rather than traditional measurements like temperature and pressure. A wide range of devices was also tested for identifying actuators that were fit to control crystal size distributions. The valuable information gathered became a guide for directing all those parties interested in crystallisation process research. This guide was updated to provide capabilities of the available measurements, as well as all possibilities for the material specific actuators.
