基于机器学习的甲基丙烯腈新工艺参数相关性研究  

Research on the Parameter Feature Correlation of Methacrylonitrile Novel Production Process Based on Machine Learning

作　　者：王昊 曲冬蕾 熊德胜 郭耘[3] 庄岩 吴君毅 蒋兆飞 WANG Hao;QU Donglei;XIONG Desheng;GUO Yun;ZHUANG Yan;WU Junyi;JIANG Zhaofei

机构地区：[1]上海华谊集团股份有限公司,上海200040 [2]上海华谊新材料有限公司,上海201108 [3]华东理工大学化学与分子工程学院,上海200237

出　　处：《上海化工》2025年第1期17-21,共5页Shanghai Chemical Industry

基　　金：上海市科技启明星扬帆项目(24YF2719000)。

摘　　要：甲基丙烯醛(MAL)氨化脱氢制备甲基丙烯腈(MAN)是一项新颖的工艺路线。针对该新型工艺路线,在前期获取的大量工艺参数调控结果的基础上,基于机器学习方法进行数据挖掘,搭建评估工作流,优化回归算法模型,并在最优模型下对各工艺参数的特征相关性进行分析。结果表明,梯度提升决策树回归及决策树回归算法是最优的回归模型,而NH3用量是影响MAL转化率及MAN选择性的主要工艺参数。The ammonification and dehydrogenation process of methylacrolein(MAL)is a novel route to prepare methacrylonitrile(MAN)which is rarely reported in previous literature.Based on the results of previous process parameter study,the data mining was operated by machine learning,including the design of evaluation workflow,the optimization of regression model,and the feature correlation analysis of different process parameters by optimization model.The results in-dicate that gradient boosted decision tree regression and decision tree regression are the optimization models,and the amount of NH3 is the most important parameter for MAL conversion rate and MAN selectivity.

关 键 词：甲基丙烯腈 工艺参数 特征相关性 机器学习 

分 类 号：TQ217[化学工程—有机化工]