Study Of An Expert System For The Formulation Design Of Osmotic Pump Of Poor Water-soluble Drug Based On Artificial Neural Network

The purpose of this thesis was to build an expert system(ES)that was mainly based on network control and production rule.The system was for the formulation design of push-pull osmotic pump tablets(PPOP)so that the efficiency of push-pull osmotic pump tablets development could be improved,the development cycle could be shorten,and the industrialization situation of osmotic pump dosage forms in China could be improved.This study was also expected to throw out a minnow to catch a whale,promote the anastomosis of disciplines,inhalans more high technology into pharmaceutics,and dedicate our strength to a faster,better and greater development of pharmaceutics and pharmacy.Water insoluble drug with extremly different dosage(1.5mg～200mg),indapamide, gliclazide and dipyridamole was employed as models.Single factor study was carried out in every dose of each drug in a large range.Over six thousand original data,eight hundred dissolution profiles and two hundred and fifty formulations were obtained through the study.The regularity of polymer,osmotic agent,binder,pore former,coat weight gain,orifice and manufacture procedure was summarized based on the data and then the database was built.Rule base of the system was built mainly based on the original data and the experiences obtained by other people before.ES is one of the most popular and active subject in artificial intelligence.No expert system available could be used for extended release dosage form design.It is gradually consummate of osmotic pump technology(OPT)no matter at home or abord. It is meaningfull to combine OPT and ES at this time.ES mainly includs man-machine interface,knowledge base(database)and inference engine.SQL Server was employed as database manage system(DBMS).Release behavior prediction model was built based on BP neural network which is good at high nonlinear mapping with learning function.Formulation design model was built based on release behavior prediction model,which was the nucleus of the inference engine.Man-machine interface was designed using VB.NET,and a computer program that could design PPOP formulation totally or partly as a professional person was come out.Finally,water insoluble drug famotidine was chosed and ES built in the study was used to design the PPOP formulations.It was considered that the ES was practicable.