Abstract: A back - propagation neural network model was used to predict the physicochemi-cal properties of organic compounds. The efficiency of linear regression and principal component analysis for determining the input parameters was demonstrated. The normal boiling points, critical temperatures, critical volumes, critical pressures, vaporization enthalpies and eccentric factor of organic compounds can be predicted quantitatively by neural network. The estimation accuracy is better enough to meet the demand for chemical engineering simulation and design.

摘要： 用神经网络法预测有机物的物性,提出用主成分分析和回归分析的方法,求得输入参数,对有机物的正常沸点、临界温度、临界压力、临界体积、汽化焓、偏心因子等基础物性作了定量预测,估算精度可满足化工计算的要求。