Pattern matching in batch bioprocesses—Comparisons across multiple products and operating conditions

A novel pattern matching method for the evaluation of multivariate time-series data is presented. The new approach allows for the comparison of batch data from industrial bioprocesses conducted at different operating conditions (or setpoints) that produce different products. By utilizing principal component analysis (PCA) and a modified PCA similarity factor (SPCAλ), comparisons of different batches using both process and quality data can be conducted. This technique results in the clustering of batches of the same product. Furthermore, comparisons between different protein products can be made. Once similarities (or dissimilarities) have been detected using SPCAλ, diagnosis steps can be taken where the relative positions of the loadings of different PCA models will determine the specific process variables that contribute to the observed phenomena. These techniques are applied to industrial data collected from a biopharmaceutical process pilot plant and show promising results for very different types of batches.