人工神经元网络在汽液物性参数拟合中的应用

Artificial Neural Networks for Parameter Fitting of Properties of Gases and Liquids

在热能系统的模拟与综合中，必须首先解决物性参数的计算问题。本文分析了现有物性参数的一些常用计算方法，提出了利用人工神经元网络对具有静态特性的物性参数进行拟合的计算机方法。在简单介绍人工神经元网络，特别是BP算法的基础上，通过对饱和水蒸汽的物性参数进行拟合的实例分析，认为人工神经元网络对于具有静态特性的饱和水蒸汽的物性参数具有很好的拟合效果，非常适合于实际的工程应用，对于其它具有静态特性的气体或液体的物性参数拟合也有参考价值。

The computing of properties of gases and liquids is most critical in the simulation and integration of heat energy systems. On the analysis of many frequently adopted fitting methods for the computation of parameters, this paper proposes a new computing method applying artificial neural networks (ANN) to fit the parameter data that show static characteristics. The principle of ANN, especially the back-propagated algorithms, is firstly introduced in brief, and then parameter data of saturated steam are fitted by the new method in case studies. As a result, applying ANN to the fitting of property parameters of saturated steam can make rather perfect effect that can completly satisfy the practical applications in engineering. And therefore, the new method can be beneficially exploited by the parameter fitting of other gases and liquids when their static characteristic parameter data are available.