Data-Driven Particle Quality Control for Spray Fluidized Bed Granulation Process by Learning from Primitive Data

Pharmaceutical quality control (PQC) holds a critical position in quality by design-based pharmaceutical development, but development costs are seldom considered due to the overly important role of drugs. Furthermore, in the context of the Made in China 2025 and Industry 4.0 strategies, drug customization brings about uncertainty embodied as frequently changing critical material attributes, which presents new challenges related to development cost and efficiency in PQC compared to traditional model-based control. As the pharmaceutical development progresses, valuable historical data, also known as primitive data, will accumulate under different raw material properties. To reduce the waste of primitive data and improve the efficiency of pharmaceutical development, this paper introduces a primitive data driven particle quality control strategy by combining case reuse and optimization-based differential compensation method, and applies it to the average particle size (APS) control in spray fluidized bed granulation process. Finally, simulation experiments are carried out from two aspects to verify and discuss the feasibility and effectiveness of proposed PQC strategies.