计算机与应用化学
計算機與應用化學
계산궤여응용화학
COMPUTERS AND APPLIED CHEMISTRY
2008年
9期
1165-1170
,共6页
合理性分析%预测模型%模型的bias%热物理性质%临界性质
閤理性分析%預測模型%模型的bias%熱物理性質%臨界性質
합이성분석%예측모형%모형적bias%열물이성질%림계성질
rationality analysis%predictive model%model bias%thermophysical property%critical property%organic compound
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,奉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槲倾向一致性的检验.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.
化學工業及科學研究極需化閤物熱力學性質的預測模型.但是現有的大多數模型可靠性很低.這主要是由于可用于髮展模型的試驗數據往往太少,以至于所用到的數據經常缺乏代錶性,使模型在作預測時易髮生大的誤差,甚至嚴重的錯誤.因此,如果數據的代錶性問題不解決,則無論任何數學模型,優化方法,神經網絡或進化算法都無法真正改進模型的預測能力,為瞭全麵地理解模型的預測能力,奉文建議除瞭要對模型作基于試驗數據的檢驗,還應對模型作基于規則的分析-即模型的閤理性分析.該分析彊調用各種結構類型的化閤物對模型預測值的閤理性作基于熱力學原理和與已知試驗數槲傾嚮一緻性的檢驗.這種分析方法不僅有助于全麵地瞭解一箇模型的預測能力,而且有助于髮展可靠的模型.
화학공업급과학연구겁수화합물열역학성질적예측모형.단시현유적대다수모형가고성흔저.저주요시유우가용우발전모형적시험수거왕왕태소,이지우소용도적수거경상결핍대표성,사모형재작예측시역발생대적오차,심지엄중적착오.인차,여과수거적대표성문제불해결,칙무론임하수학모형,우화방법,신경망락혹진화산법도무법진정개진모형적예측능력,위료전면지리해모형적예측능력,봉문건의제료요대모형작기우시험수거적검험,환응대모형작기우규칙적분석-즉모형적합이성분석.해분석강조용각충결구류형적화합물대모형예측치적합이성작기우열역학원리화여이지시험수곡경향일치성적검험.저충분석방법불부유조우전면지료해일개모형적예측능력,이차유조우발전가고적모형.
Chemical industry and scientific studies depend heavily on predictive models of thermophysical properties.Unfortunately, the reliability of many models is extremely limited.This is mainly because the available experimental data for model developments are often too few to be representative of the compounds interested by researchers.Thus, the reliability issue of models cannot be resolved no matter what kinds of mathematical models, optimization methods, neural network or evolutionary algorithms are adopted if the structural features of sample compounds are limited.Recognizing this limitation is crucial for model users to appropriately use the predicted data.This report suggests that for a full understanding of a model reliability in prediction, in addition to an experimental data based evaluation, it is also necessary to perform a rule-based rationality analysis on the model.This analysis should be conducted by systematically examining the rationality of values predicted by the model with a set of rules, which consist of thermodynamic principles and experimental data characteristics for a given property.By this analysis, compounds with a broad diversity in structure should be tested, even when there are no observed data for these compounds.
