Research Of Data-driven Quality Control Method In Drug Fluidized Bed Granulation Process

The fluidized bed granulation,process(FBGP)is an important part of the drug production process.During drug development phase,the frequent adjustment of drug formulation gives rise to the change of raw material properties and quality indicators,which leads to difficulties in the quality control based on the model-based control approach.The data-driven quality control is to directly design the controller by using the input and output data of the process.Therefore,it is of great theoretical and practical significance to study the data-driven quality control of FBGP.This thesis first introduced the basic principle,process flow and mechanism model of FBGP,and verified the effectiveness of mechanism model through simulation experiments.Then,the basic concepts and basic ideas of the data-driven control(DDC)were described.At the same time,two DDC methods,namely model-free adaptive control(MFAC)and data-driven optimal iterative learning control(DDOILC)were introduced and applied to the particle quality control of FBGP.Their control performance was analyzed by simulation studies.On this basis,according to the respective advantages of MFAC and DDOILC,two improved data-driven model-free adaptive control(DDMFAC)methods were proposed.The first one was to combine the control objectives of MFAC and DDOILC to improve the control performance according to the performance characteristics of the two control methods when designing the DDMFAC controller.The second was to determine the selection of MFAC and DDOILC based on the error information between system output and the desired value and the related method parameters were tuned by the fuzzy controller.Then,the convergence problem of the above improved control methods was discussed,and the control performance of the two improved methods were compared by simulation.Finally,using the historical data and operational experience knowledge under the conditions of different raw material properties and quality indicators accumulated with drug research and development,and combining real-time modeling and case-based decision-making methods,the data-driven particle quality control strategy of FBGP based on primitives was proposed.The simulation results verified the effectiveness of the proposed strategy.