Abstract The relationship between flammability limits (FL) of gas mixtures and their molecular structures was studied in this research based on the quantitative structure-property relationship (QSPR) method. The genetic algorithm (GA) was used to select an optimal subset of descriptors which show a significant contribution to the properties. The external validation, which checking the stability and predictive capability of the obtained models, was focus of this study. Three different external validations were employed to build the models. Moreover, the applicability domain for the models was also defined. The selected descriptors and obtained models were then extended to predict the FL values of existing or new mixtures based only on mixtures composition and their molecular structures.

中文翻译：

---

一种预测气体混合物可燃极限的数学方法

摘要 本研究基于定量结构-性能关系（QSPR）方法研究了混合气体的可燃极限（FL）与其分子结构之间的关系。遗传算法 (GA) 用于选择对属性有显着贡献的描述符的最佳子集。检查所获得模型的稳定性和预测能力的外部验证是本研究的重点。使用三种不同的外部验证来构建模型。此外，还定义了模型的适用范围。然后将选定的描述符和获得的模型扩展到仅基于混合物组成及其分子结构来预测现有或新混合物的 FL 值。