Artificial intelligence infrastructure into material attributes insights

Abstract

The development of a biopharmaceutical manufacturing process involves an assessment of all possible sources of variation throughout each of the unit operations in the drive toward six sigma manufacturing. The primary goal of this project is to develop a novel way to assess the variation in raw materials attributes throughout the life-cycle of the material and gain insights about the correlation between material variation to process performance and product quality. This thesis focuses on understanding the impact raw materials have on unit operations within biopharmaceutical manufacturing processes through machine learning techniques. To evaluate the impact of raw material attributes on process performance and exclude the variations explained by process operating parameters, a modeling framework is developed and tested. The framework contains three steps: (1) fitting models with only process operating data, (2) fitting models with process operating data and batch number information, (3) fitting models with process operating and raw material attributes data. By comparing the performance measurements from 3 different models, insights of correlations between raw materials and process outcomes could be obtained.