Prediction of batch quality indices using functional space approximation and partial least squares

Due to regulations and mandates in the pharmaceutical industry, it is important to be able to achieve real time release and achieve consistent batch runs. From a batch reporting perspective as well as for proactive scheduling of downstream steps in the manufacturing, accurate predictions of the end point quality indices of the batch is important. This paper presents an approach that is based on functional space approximation and multiway PLS for prediction of the batch quality indices. A novel yet simple method for completing a batch record during online batch monitoring and prediction is proposed. The proposed methodology has been validated on representative simulations involving a fed-batch fermentation for the prediction of final antibiotic yield as well as the batch durations.