Data-driven Particle Quality Control of Spray Fluidized Bed Granulation Process

Spray fluidized bed granulation (SFBG) is a widely applicable process in the pharmaceutical industry, and the particle quality control is of great significance both theoretically and practically. In the drug development phase, however, frequent adjustment of prescription will give rise to the variation of material properties, which will lead to the different initial values (IVs) of particle quality for the same set of empirical operating condition. Most of all, frequent adjustment of prescription will result in deficiency of historical data and operational experience for new prescription. Because it is laborious and resource-wasting to get precise process model whenever the prescription changes, so that model-based control (MBC) method is no longer applicable. Therefore, A data-driven model-free adaptive control (DDMFAC) strategy is proposed by combining the advantages of model-free adaptive control (MFAC) and data-driven optimal iterative learning control (DDOILC) to achieve particle quality control of SFBG. Firstly, the influences of algorithm parameters on control performances of MFAC and DDOILC are studied to explore the parameter setting rules of the control process. Then, Combined with the previous theoretical analysis for MFAC and DDOILC, the advantages of MFAC and DDOILC are separately reflected in different control stages to achieve the better control performance. Lastly, the fuzzy adjustment strategy is utilized to select suitable method (MFAC or DDOILC) and appropriate algorithm parameters in different control stages. A series of simulation experiments verify the effectiveness of proposed DDMFAC strategy.