This case study compares using OSIsoft PI Asset Framework and Event Frames versus manual techniques to aggregate biopharmaceutical batch data and test attributes from disparate data source systems to enable multivariate statistical modeling. The primary objective is comprehensible integration of batch process data with recipe conditions, test results, raw material attributes, and process analytical technology to model the evolution of a batch, including predictive and prescriptive analytics. Various data science methods are achieved by wrapping contextualized time-resolved batch event data for analysis, visualization and reporting.
Interface, asset framework, AF, event frames, EF, data warehousing, link tables, cycle times, PI DataLink, principal components analysis, PCA, partition of variance components, projected latent structures, PLS, batch evolution modeling, SIMCA-online, univariate bivariate and multivariate statistics.
Will A. Penland, MSPA is the Principal Data Scientist, Data Analytics & Operations Excellence at Boehringer Ingelheim, Animal Health in St. Joseph, MO.