(460d) On-LINE Optimal Control of Crystal Properties in NON-Isothermal Antisolvent Crystallization

Controlling crystal properties has been a challenging issue in many crystallization operations due to inevitable disturbances, occuring during the process. This contribution proposes an online strategy to control crystal size distribution in joint cooling antisolvent crystallization systems.

As a first step in the study, digital image texturing analysis in the form of fractal dimension and energy signatures is applied and automated to track crystals’ properties at different time of the experiment. The one dimensional Fokker-Planck equation (FPE) represents the dynamic characteristics of the crystal growth. A map of iso-mean and iso-variance curves are reported in an antisolvent flow rate-temperature plane based on evaluating the model at asymptotic condition. This allows identifying the region of input multiplicities in the system so punctual information regarding stability of the process can be obtained. Using controllability tools, regions of singularities needed to be avoided for proper control of the process is determined and it is further shown that the system is ill conditioned in the whole operating map. The problem is investigated experimentally where a multi-loop PID structure is tuned by pairing crystals’ mean size with antisolvent feed rate and manipulating temperature to control the size variation. The proposed controller is demonstrated to be incapable of bringing the system into the desired set points and an offset in both crystals’ mean size and variance is observed at the end of the batch. In the next step, a feed forward approach is developed in which temperature and flow rate trajectories are obtained from a static optimization procedure. The sensitivity of feed forward control strategy is assessed in the case of disturbances in antisolvent feed rate. Finally, in order to achieve a better performance a two stage controller is set up by combining the feed forward approach with a feedback controller which is designed based on IMC (Internal Model Control) algorithm. It is indicated that a perfect control over crystals’ mean size can be obtained in this case by renouncing a strict control over size distribution coefficient of variation. Experimental validation of the strategies are carried out for the ternary system of water-ethanol-sodium chloride.