Construction Of Automatic Control System For Crystallization Process And Product Quality Control Of Aspirin

Crystallization is an important unit operation in the production of drugs.More than 90% of the active pharmaceutical ingredients(APIs)exist in the crystalline form.Since the crystal mean size and size distribution affect the stability and flowability of products significantly,the crystal size distribution(CSD)is the key objective of crystallization optimization and control.The previous investigations for optimization of crystallization were mostly based on design of experiment(DoE)and trial-and-error,which could not excavate the essence of the process quickly and deeply.With the development of process analytical technology(PAT),the measurement of multiple process variables and crystal properties in real time has been possible.It has become an important tool for process control.In this study,a crystallization process automatic control system(CPACS)was developed by integrating PATs,data mining and processing,with several control strategies including PID and population balance(PB)model.The spectrum and temperature signal were collected and analysed with chemometrics model in real time.Then the evolution of virtual variables such as supersaturation would be predicted,and it could improve the understanding of the crystallization process from the perspective of multiple variables.Furthermore,the control strategies of crystal size based on the constant supersaturation and population balance model(PB)were also studied in this thesis.Specifically,the crystallization of aspirin in ethanol was investigated to evaluate the ability of CPACS in process monitoring,data processing and correlation,and the control of CSD.The main contents are as following:(1)The ultraviolet spectra of aspirin-ethanol system were collected in real time.The prediction model according to partial least squares(PLS)method was established and the characteristic wavelengths were optimized by genetic algorithm.Based on the module,CPACS was able to measure the solubility of aspirin in ethanol automatically.The average relative error of solubility data was 3.28% comparing with the gravimetric method.Then the data was fitted with the classical thermodynamic equations and the correlation coefficients were all larger than 0.9991.(2)The metastable zone width was measured by CPACS.It shows that CPACS is more sensitive than the turbidity method.The result demonstrates that the metastable zone width becomes much wider as the increasing cooling rate or the reducing stirring rate.(3)On the basis of PB model and mass conservation equation,the evolution of solution concentration and crystal size were simulated.The residual sum of squares of the simulated concentration and CPACS predicted values was regarded as the objective function and subsequently the parameters of crystal growth rate were estimated.Furthermore,the results were validated with the stereo imaging.It indicates that the orders of magnitude of the two methods are the same,and the average relative error is 13.55%.(4)PID was used to keep the supersaturation at a constant value and to obtain products with much narrower CSD.Additionally,the optimized curve of cooling rate to get the products with mean size of 350 ?m,was obtained by optimizing PB model with genetic algorithm.Finally,the optimum cooling curve was used to control the crystallization process,and the products with the mean size of 347 ?m were obtained.