In this study, we introduced an artificial neural network (ANN), which can represent objects that are difficult to formulate, rather than the integrated model of the group contribution method (GCM), because the integrated model in the existing GCM is no longer able to manage enormous diversification of substances. Hence, a model to estimate the pure component parameters m, σ, and ε of the perturbed chain statistical associating fluid theory (PC-SAFT) equation of state (EoS) was developed. We optimized the structure of the ANN by changing the number of neurons and layers in the hidden layer. In this study, the optimized ANN structure was two hidden layers and 40 neurons. The results confirm that the model can determine the pure component parameters of PC-SAFT EoS, which can estimate the liquid density, saturated vapor pressure, and critical properties. In terms of the critical properties, the estimated results were almost as good or better than those obtained using the GCMs, which are specified for the critical properties. We were able to calculate properties for substances whose parameters were not reported in the literature by using this ANN. The results would be useful for chemical process design, for example, to be incorporated into process simulators.

在这项研究中，我们引入了人工神经网络（ANN），它可以表示难以制定的对象，而不是群贡献法（GCM）的集成模型，因为现有 GCM 中的集成模型不再能够管理物质的巨大多样化。因此，用于估计纯分量参数m、σ和ε 的模型扰动链统计相关流体理论 (PC-SAFT) 状态方程 (EoS) 的发展。我们通过改变隐藏层中的神经元和层数来优化 ANN 的结构。在本研究中，优化的 ANN 结构是两个隐藏层和 40 个神经元。结果证实，该模型可以确定PC-SAFT EoS的纯组分参数，可以估计液体密度、饱和蒸气压和临界性质。在关键特性方面，估计结果几乎与使用 GCM 获得的结果一样好或更好，后者是为关键特性指定的。通过使用该人工神经网络，我们能够计算文献中未报告参数的物质的特性。结果将有助于化学工艺设计，例如，