【摘要】 乙醇-环己烷二组分系统的气液平衡相图实验是一个重要的本科物理化学实验，然而其传统教学方案存在结论简单、内容单一等局限。本研究采用机器学习方法对该实验进行了数智化升级。通过在9 124个不同组分的气液相平衡数据上训练的随机森林模型，实现了乙醇-环己烷及其衍生物体系气液平衡相图的准确预测，并利用SHAP分析揭示了影响相图形态的关键因素是二组分的沸点之差。一轮教学实践结果表明，本数智化设计方案不仅加深了学生对气液平衡相图规律的理解，而且通过数据分析和模型解释，为学生探索不同形态的相图提供了一个交互式平台，增强了课程的吸引力。该研究促进了数智化技术与化学实验教学的深度融合，有望在实验教学领域发挥示范作用。更多还原

【Abstract】 The gas-liquid equilibrium phase diagram determination experiment of ethanol-cyclohexane two-component system is an important undergraduate physical chemistry experiment.However, the traditional teaching program has limitations due to the simple conclusions and content.In this study, machine learning approaches were used to upgrade this experiment with digitalization and intelligence.The random forest model trained on 9124 different components of the gas-liquid equilibrium dataset achieves accurate prediction of the gas-liquid equilibrium phase diagram of ethanol-cyclohexane and its derivatives.SHAP analysis reveals that the key factor affecting the morphology of the phase diagram is the boiling point difference of the two components.A round of teaching practice shows that the digitized and intelligent design scheme not only deepened students’ understanding of gas-liquid equilibrium phase diagram, but also provided an interactive platform for students to explore different forms of phase diagram through data analysis and model interpretation, and enhances the attractiveness of the course.This research promoted the deep integration of data intelligence technology and chemical experimental teaching, and was expected to play a demonstration role in the field of experimental teaching.更多还原