论文部分内容阅读
化学工业及科学研究极需化合物热力学性质的预测模型.但是现有的大多数模型可靠性很低.这主要是由于可用于发展模型的试验数据往往太少,以至于所用到的数据经常缺乏代表性,使模型在作预测时易发生大的误差,甚至严重的错误.因此,如果数据的代表性问题不解决,则无论任何数学模型,优化方法,神经网络或进化算法都无法真正改进模型的预测能力.为了全面地理解模型的预测能力,本文建议除了要对模型作基于试验数据的检验,还应对模型作基于规则的分析-即模型的合理性分析.该分析强调用各种结构类型的化合物对模型预测值的合理性作基于热力学原理和与已知试验数据倾向一致性的检验.这种分析方法不仅有助于全面地了解一个模型的预测能力。而且有助于发展可靠的模型.
Chemical industry and scientific research require predictive models of the thermodynamic properties of compounds, but most of the current models are less reliable, mainly because of the paucity of experimental data available to develop the models so often that the data used are often poorly represented Therefore, if the representativeness of the data is not solved, no mathematical model, optimization method, neural network or evolutionary algorithm can really improve the model’s Predictive ability.In order to fully understand the predictive ability of the model, this paper suggests that in addition to test the model based on the test data, the rule-based analysis of the model, that is, the rationality analysis of the model, should be emphasized.This analysis emphasizes that using various structural types The rationale for compounds’ predictive value for the model is based on the thermodynamic principles and tests of agreement with the known test data, which not only helps to fully understand the predictive power of a model. But also help to develop a reliable model.