(654c) Experiences in the Synchronization of Batch Trajectories for Pharmaceutical Process Analysis and Monitoring | AIChE

(654c) Experiences in the Synchronization of Batch Trajectories for Pharmaceutical Process Analysis and Monitoring


Polizzi, M. - Presenter, Pfizer Worldwide Research and Development
Prpich, A. - Presenter, Pfizer Global Research & Development
Strain, C. - Presenter, Pfizer Global Research & Development
Lalonde, A. - Presenter, Pfizer Global Research & Development
Negron, V. - Presenter, Pfizer Global Manufacturing

When the dynamic behavior of the data from a batch process is analyzed (irrespectively of the unfolding approach) the data needs to be ordered such that measurements from different evolution points are not mixed into a diagnostic. This pre-processing step is known as trajectory alignment, which is a necessary step in the application of multivariate analysis of batch data. Multiple techniques are found in literature to deal with this problem; from the simple re-sampling against an indicator variable [1], to more complex algorithms like Dynamic Time Warping [2,3].

In the case of a pharmaceutical product and the regulations involved with such an operation, operation recipes are always found for a process. These recipes are mostly written around triggers in the operation that serve as indicator variables for alignment. This work presents our experience in performing this alignment with various processes (from drug product and drug substance) and the found advantages for each in developing monitoring schemes or soft sensors for product quality prediction.

It also presents a quick and approximate method to triage a new set of data to determine if the analysis of its dynamic features should be progressed or not. This method is based on the analysis of the importance of the off-diagonals of the variance-covariance matrix of the scores from an observation-wise unfolded PCA model, from which a PLS model fitted against final product properties or quality classification. The new method is illustrated with a drug product process.

Literature Cited

(1) Nomikos, P. ; MacGregor, J. F. Monitoring Batch Processes using Multiway Principal Components. AICHE J. 1994, 40 (8), 1361-1375. (2) Westerhuis, J.; Kourti, T.; Kassidas, A.; Taylor, P.; and MacGregor, J. F. Synchronizing the trajectories of the process variables for on-line monitoring of batch runs with unequal duration. SSC6, Porsgrunn, Norway, August 15- 19 1999. (3) Kassidas, A.; MacGregor, J. F.; and Taylor, P. Synchronization of batch trajectories using dynamic time warping. AICHE J. 1998, 44 (4), 864-875.