Monitoring and Control of Crystal Shape in Crystallization Processes

Publication Reference: 
Author Last Name: 
James B. Rawlings, Daniel B. Patience
Report Type: 
FRR - Final Report
Research Area: 
Particle Formation
Publication Year: 
Publication Month: 
United States

In this research, we successfully implemented feedback control of particle shape in a semi-batch crystallization. The overall goal of this research was to measure and regulate the shape and size of particles created by nucleation and growth processes in crystallizers. The state of the art in this field up to 1993 is summarized in the review article [12]. At that time, control of crystal size and crystal size distribution was just becoming possible using simple on-line slurry measurements such as light transmittance or small angle forward light scattering. The challenge undertaken in this research was to go to the next level and attempt direct control of crystal shape. It was felt that demonstration of online shape measurement and control would require the development of entirely new measurement technology compared to what was being used for size control.

The measurement technology we developed for this purpose was direct digital imaging of a sample stream. As discussed in the report, the key to extracting useful particle shape information from the digital images requires the user to monitor two key variables. We chose boxed area and aspect ratio of the identified particle images to infer the shape of the crystals.

We developed the following crystallization system to demonstrate the result. An impurity free stream flowed through the crystallizer and we regulated the flowrate of a habit modifier stream in order to maintain the desired shape. At the 2000 IFPRI Annual meeting, we showed our first results in which, without any prior knowledge of model parameters, a simple proportional-integral control algorithm is able to maintain a desired crystal shape and in doing so, determines the critical concentration of habit modifier required to maintain this shape.

The prototypical system and process we selected is semi-batch crystallization of sodium chlorate (NaClO3 ). Sodium dithionate (Na2 S2 O6 ) is a habit modifier that influences the  relative growth rates of 100 and 1 ̄1 ̄1 ̄ faces of the crystal. In the presence of at least 50 ppm sodium dithionate the growth of the 1 ̄1 ̄1 ̄ faces is blocked by the impurity and the crystal shape changes from cubic to tetrahedral. Without impurity present, the 100 faces grow slower than the 1 ̄1 ̄1 ̄ faces and the crystal shape changes from tetrahedral to cubic. The shape change is easy to detect with video images alone, though there are limitations with extracting useful quantitative information from images for use as a signal for feedback control.

This prototypical process displays the following industrially relevant characteristics.

1. Particle shape is affected by unmeasured disturbance variables.

2. Online sensing is available in the form of video images. The images are replete with bad data. Some particles are fused or broken; it is difficult to obtain representative samples; particle boundaries overlap each other; there are significant levels of process noise; and it is difficult to sample enough images to remove the effects of this noise through averaging. The standard image analysis software provides simple measures such as particle boxed area and aspect ratio; as we show later, these simple measures are inadequate signals for feedback control.

3. We can manipulate a process variable that also influences particle shape. Through this feedback policy, we maintain the desired shape in the face of the unmeasured disturbances. The video images are processed in real time to produce the feedback signal that is used for control.