Analytical needs for continuous processing, a shift from assurance to automation and advanced control
Time: 15.00-16.00 (CET)
Sartorius Stedim Data Analytics (SSDA) invites to a webinar on the use of the Umetrics® Suite in continuous manufacturing
The history of regulatory influence and prevalence of batch type processing in the manufacturing of pharmaceutical products has given rise to a culture driven by quality assurance. As a result, analytics developed to monitor and control these processes focused on identifying risk of producing poor quality product. The analytical needs for continuous processes are unique relative to batch. There is a shift away from monitoring, towards automation and advanced control. The automation systems for continuous processes are designed to respond to variations in raw material and measurements of intermediate and final qualities. The manufacturing system is inherently dynamic, continually adjusting its operating targets to achieve a consistent quality. This is in contrast to traditional batch manufacturing in health science where the objective is to execute a consistent operating policy or recipe. Recent initiatives in the regulatory environment towards science based quality systems has helped in the adoption of more modern automation systems, but this transition is slow moving and still quite limited.
In this talk a review of typical methods used in industries such as petrochemical and foods is highlighted. Automation methods such as model predictive control (MPC) are described, including a discussion on integration with quality control/assurance. Relative to batch manufacturing, quality assessment in continuous manufacturing requires the use and understanding of analytics that address the unique challenges of continuous flow. Non-ideal behavior from axial mixing, back mixing and volume buffering leads to a time-distributed influence of process variation on material qualities. It is shown how application of multivariate (MV) data unfolding methods can be used to address coordination of multiple continuous unit operations and the respective residence time distributions (RTD) from each. Analogies are provided demonstrating the similarity and difference of these methods to common multivariate methods used to monitor and predict batch process performance.
Finally state observers are presented. State observers are similar to traditional soft sensors in that they provide estimates of hard or slow to measure attributes. The unique feature of state observers is they are built on mechanistic understanding of the process kinetics. These are tools that bridge the gap between purely data driven methods and mechanistic understanding.
Click here to register