(207e) Modeling, Simulation and Control of Twin-Screw-Granulation and Fluid-Bed-Drying Applied to a ConsiGmaTM-25 Line
AIChE Annual Meeting
Monday, November 14, 2022 - 4:54pm to 5:15pm
An essential component for the development of such control concepts is the real-time monitoring of critical quality attributes (CQA). The granule size after the TSG is considered to be an intermediate CQA. Therefore, a Parsum probe, capturing the particle size distribution (PSD) of wet granules, is placed at the TSG outlet. From this distribution data, relevant granule size characteristics are computed. A process model describing the relation between the granulation parameters and these characteristics has been developed by means of a local-linear model tree (LoLiMoT) algorithm. Model predictive control (MPC) uses this model as a core component and selects optimal granulation parameters in order to produce granules of well-defined PSD. The concentration of the active pharmaceutical ingredient (API) is another important CQA. This quantity is acquired by means of a Raman probe mounted after the TSG in combination with a model that computes the API concentration from the Raman spectral data. This information is fed to a feedback control concept that adjusts the process parameters in order to keep the API concentration close to its reference value. In parallel, the developed PSD model is utilized as a soft-sensor, allowing the real-time granule size monitoring, and acting as a potential replacement for the Parsum probe. Granule moisture after the FBD is a further CQA that should be monitored in real-time. For that purpose, a mathematical model based on physical laws has been developed and parametrized on the experimental data. This model runs in parallel with the process, estimating the granule moisture from the collected process data. Based on this estimation, a control concept stops the drying as soon as the desired moisture value has been reached. The performance of introduced control concepts has been examined on the ConsiGmaTM-25 plant by means of selected disturbance scenarios.