Machine vision and image processing for intelligent online monitoring and control of pharmaceutical fluidized bed granulation processes

Fluidized bed granulation is a unit operation widely used in the pharmaceutical, chemical and food processing industries. It is a manufacturing technology that by suspending lose powders using hot air and transforms the powders into granules of uniform sizes to improve compaction and flow characteristics. The granule size distribution and moisture content are important quality indicators that are currently characterized by sampling and offline analysis in the laboratory, leading to time delay in measurement. This work reports an investigation of machine vision combined with deep learning image segmentation for on-line real-time monitoring. A non-invasive microscopic imaging probe with an integrated light source is designed and mounted on the granulator’s sight glass to monitoring the granule dynamic changes in particle morphology and size. In addition, a near-infrared spectrometer combined with chemometric modeling is used for real-time monitoring of moisture content.