 |
Improve Crystallizer Uptime |
This application note describes how a challenging crystallization process, designed to produce an a high-value chemical, can be optimized using an automated feedback loop. The crystallization process under investigation sometimes relies on incoming raw materials with varying impurity profiles and may produce a non-stirrable suspension, requiring crystallizer shutdown to perform a manual isolation; the cost associated with this shutdown, and the subsequent cleaning procedure can easily reach extra costs of €50,000 per batch, or even more. In order to address this issue, a self-seeding strategy was designed whereby spontaneously nucleated crystals were refined in situ through temperature cycling. A straightforward, feedback control loop was set up to prevent full seed dissolution in a challenging batch environment, with frequently changing impurity levels and compound solubility temperatures. A scale-down study utilizing this self-seeding strategy successfully demonstrates controlled secondary nucleation and desired crystal growth, indicating improved uptime and reduced cost is feasible during production.
Contents
1. Introduction
2. Process Understanding for a Challenging Crystallization
3. Development of a Self-Seeding Strategy Under Varying Impurity Conditions
4. Automated Feedback Control for Optimized Operation at Unknown Purity Level
5. Summary
6. Appendix
In the agrochemical industry, the development of automated production processes receives increasing attention in order to continuously improve manufacturing efficiency. The key target (a final product within specification) needs to be met at maximum throughput, reduced consumption of raw materials, and in a robust and repeatable way. Any disturbance to this process can lead to poor filtration performance, crystal rework requirements, lengthy cleaning procedures, and even plant shutdown, which are all associated with significant undesirable extra cost.
For crystallization processes, crystal size, shape, and concentration are key parameters to ensure smooth operation. It is vital to routinely study crystallization mechanisms during process development and optimization. However, for most crystallization processes, this is not usually done. Instead, process understanding is obtained based on sampling and endpoint offline analysis. Representative sampling is often difficult or impossible, and once the offline results are available, they are often out of date because the state of the crystallization has already progressed in the meantime. Challenges are traditionally addressed in a trial and error approach without understanding the root cause, and once a presumably good process is identified, the difficulty of scale-up remains. In order to improve production uptime, it is important to be able to run processes repeatedly and robustly, even under conditions of changing vessel and stirrer geometries, heat exchange area, or slightly changing impurity levels in the raw material. Traditional methods, such as sampling and offline analysis, are usually good options for final product quality control, but the long lead time and the large number of samples required for comprehensive process monitoring makes them unsuitable to detect unplanned operations quickly. Basic in situ probes, such as turbidity sensors, are very good at showing the onset of phase transformation but can neither distinguish between oiling out and nucleation, nor are they sensitive enough to clearly deconvolute growth and agglomeration. Turbidity probes always need to be calibrated to a specific system, and do not provide size or count information, which is particularly important when working with seeded processes to constantly ensure the required seed size and number.
A solution to the process understanding challenges during development is to apply real-time microscopy, which provides unambiguous real-time in situ process images of particle mechanisms as they happen in the process. For a more detailed analysis during process optimization, scale-up, and for robust plant operations, an inline calibration-free particle size and count analyzer allows identification and fingerprinting of the ideal process. Through a feedback control loop, it can automatically react to undesirable particle mechanisms in real time.
