【摘要】 溶解度的测定与预测对于多晶型体的晶体生长和结晶过程中的多晶型控制至关重要.利用激光监视装置,首次测得了半水盐酸帕罗西汀在不同溶剂体系中的溶解度,共计308组数据;采用多参数人工神经网络模型,随机选取308组数据中的184组数据进行人工神经网络的训练,考察了不同隐含层节点数对神经网络训练效果的影响,得到了优化后的人工神经网络模型,利用剩余的124组数据对优化后的模型进行检测,平均预测误差小于0.7%.预测结果表明,优化后的人工神经网络模型可以胜任半水盐酸帕罗西汀溶解度预测的任务.更多还原

【Abstract】 Modeling and prediction of solubility is a key to develop polymorph crystal growth and crystallization process. In this paper,by using a laser monitoring observation technique, solubilities of paroxetine hydrochloride hemihydrate in various solvents were determined by the synthetic method and a multiparameter artificial neural network (ANN) approach was used to predict the solubility of solute in different solvents. The ANN predicted the logarithmic solubility of paroxetine hydrochloride hemihydrate with average error of 0.7%. The results showed that artificial neural network could be useful when no data is available.更多还原