Using Feed-forward Neural Network Predict Pure Saturated Liquid's Physical Property

Saturated liquid's thermodynamics and transfer property are indispensable to chemical engineering, especially the relevancy of physical property and temperature is important in chemical research and production. At present ,because there are many shortages in various estimated methods ,using Artificial Neural Network (ANN) to predict relationship of physical property and temperature are present in this paper.In this paper, we study learning method of using ANN model to predict relationship of physical property and temperature . Choosing universal three layer feed-forward ANN model as work network, putting forward an improved learning arithmetic viz. Damped Newton Method .at the same time we study measures to improve precision. experiment results show: lessening non-liner extent of function; choosing appropriate work range; appropriate middle cell number and systematically organizing data can quicken convergence of network and improve estimation.Using ANN to estimate thermodynamics property of saturated liquid, and receive good result, using trained network to predict stylebook , the average estimated error of heat capacity.. heat of evaporation-, density and steam pressure are respectively 0.0123%-. 0.018%-. 0.022%. 0.125%.We study relationship of transfer property and temperature further, for the data of stylebook, the average estimated error of coefficient of heat conductivity-. surface tension.. viscosity are respectively 0.089%-. 0.037%-. 0.125%.comparing the estimated data with the original literate data, they are nearly identical .this result shows that it is successful to using ANN to predict the physical property.At last, in order to convenience query and estimation, we develop Petroleum Chemical Physical Property Estimated System (PCPPES). User can query and estimate physical property according to needs.