Hardware and assembly issues in centrifugation can cause product loss, improper separation, and cell lysis. These issues have cost up to $1 million per batch in a large scale biopharmaceutical manufacturing process. If we could predict and fix these types of problems before they happened, we would have more robust and reliable manufacturing processes. Recent advances in process informatics have made a lot more process and equipment data much more accessible. A number of mathematical methods were used to analyze the clean-in-place (CIP) process data collected before each batch. We were able to distinguish all 5 bad batches from total about 80 batches in an historical dataset. Among the methods, Principle Component Analysis is the most promising and is able to detect issues in 10 subsequent batches. The methods are now being integrated into a real-time application that will warn operators to stop the operation before the batch starts if a problem is predicted.

Prediction of subsequent centrifuge problems by analyzing clean-in-place data

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.