Application Of NIRS For Process Intelligent Monitoring And Controlling In Fluid Bed Granulation

The fluidized bed is a semi-continuous production instrument,which integrating mixing,granulation and drying.Generally,the CQAs(moisture,particle size,and API content uniformity)were determined using offline analysis methods,which cannot reflect the physical and chemical properties of the materials in the fluid bed.The quality control method lags behind the production process and cannot adjust the fluidized bed process parameters in real time according to changes in the quality of raw materials and environmental factors,resulting in poor consistency between granules.Near Infrared Spectroscopy(NIRS),as the most common process analysis technology(PAT),is widely used to monitor the changes of CQAs in the pharmaceutical manufacturing process.In this study,a portable near-infrared spectrometer,combined with chemometric methods and control algorithms,was used to monitor and control the CQAs(moisture,particle size,drug content uniformity)in fluidized bed granulation process.The main purpose was to monitor and control production process of fluidized bed granulation,realize the concept of "Quality by Design(QbD)" and ensure product quality.Then a set of intelligent quality control key technology can be formed,which will provide reference and technical means for the quality management of the whole solid pharmaceutical production process.The specific research content of this paper was as follows:(1)Application of NIRS in monitoring API content of fluid bed granulation.In this experiment,a portable near-infrared spectrometer was used to collect inline spectra of the fluidized bed granulation and drying process with an API formula change of 75-125%.The API content model was established using and an API prediction model was established using chemometric methods.The influence of formula factors on the robustness of EIOT in the prediction of API content was investigated through experimental design.The prediction performance of two algorithm,extended iterative optimization technology(EIOT)and partial least squares(PLS),was compared.The experimental results show that the prediction performance of the EIOT method was better than that of the PLS method.The API content prediction accuracy of the EIOT method was verified through offline HPLC measurement.At the same time,the spatial distribution of API was measured using Raman imaging technology.The imaging results show that the API was evenly distributed on the granules,which further verified the accuracy of the EIOT method.(2)Feasibility study of intelligent control of fluidized bed granulation process based on NIRS.The PLS model of moisture and particle size was established,whose robustness was verified through investigating the process and formula factors using experimental design.Then the established model and the PID control algorithm were used together to monitor and feedback control the moisture in fluid bed granulation.The experiment investigated the feasibility of two control methods(constant moisture control and gradient moisture control).The optimal PID parameters two methods were optimized through the engineering experiment,which was verified by error analysis.(3)Study on the Evaluation and Robustness of the Control Model of fluid bed granulation process based on NIRS.This experiment compared the end-point particle distribution of moisture gradient control granulation(MGCG),moisture constant control granulation(MCCG),and regular granulation(RG).The consistency of moisture gradient control granules was obviously better than that of two other granulation methods,which was proved by measuring the relative particle size distribution(RW)and RSD(D10,D50,D90)of three granulation methods.The result of PCA showed that the moisture-controlled granulation was obviously different from the regular granulation.The influence of formula factors(binder ratio)and process parameters(temperature,atomization pressure)on the performance of gradient moisture control was further examined through experimental design.The actual value of moisture can reach the set value quickly,and the repeatability of particle size production curve between different batches under the same control condition was very well,which proved the robustness of the control algorithm.