摘要
氢化丁腈橡胶(HNBR)力学性能与橡胶配方和加工工艺密切相关.为探究材料配方与工艺对氢化丁腈橡胶力学性能的影响规律,笔者收集了32篇公开报道文献中的313份实验研究数据,提取了各文献中的体系配方、硫化工艺、橡胶拉伸强度数据,设计了极端梯度提升模型(XGBoost)与类别增强型提升模型(CatBoost)2种机器学习模型.首先对输入特征进行独热编码,之后采用2种机器学习方法进行训练,比较2种模型的预测精度、泛化能力,并进行特征重要性分析.2种模型的预测精度均超过0.92.特征重要性分析表明,炭黑含量和交联剂含量为关键的工艺参数,但2种模型描述的特征重要性比率存在差异.研究结果对研究氢化丁腈橡胶的工艺配方设计和发展机器学习技术在橡胶材料领域的应用具有重要的探索意义.
The mechanical properties of hydrogenated nitrile butadiene rubber(HNBR)are closely related to the rubber formulation and processing techniques.To investigate the influence of material composition and processing on the mechanical properties of HNBR,the authors collected 313 publicly reported research studies in 32 papers,extracting formulation systems,vulcanization processes,and tensile strength data from the literature.Two machine learning models,extreme gradient boosting(XGBoost)and categorical boosting(CatBoost),were developed.The input features were first one-hot encoded,followed by training using both models to compare prediction accuracy,generalization ability,and feature importance analysis.Both models achieved a prediction accuracy of over 0.92.Feature importance analysis showed that carbon black content and crosslinking agent content are key processing parameters,though the importance ratios differed between the models.The results could provide valuable insights for the design of HNBR formulations and the application of machine learning techniques in the rubber materials field.
作者
丁瀚林
赵骞
张洁
孙思嘉
陈皓哲
陈鹏
DING Hanlin;ZHAO Qian;ZHANG Jie;SUN Sijia;CHEN Haozhe;CHEN Peng(Anhui Provincial Key Laboratory of Green Polymer Materials,School of Chemistry and Chemical Engineering,Anhui University,Hefei 23060l,China;Electric Power Research Institute,State Grid Anhui Electric Power Co.,Ltd.,Hefei 23060l,China;Faculty of Science and Engineering,University of Nottingham,Ningbo 315100,China)
出处
《安徽大学学报(自然科学版)》
北大核心
2025年第3期90-99,共10页
Journal of Anhui University(Natural Science Edition)
基金
国网安徽省电力有限公司科技项目(521205220003)。
关键词
氢化丁腈橡胶
机器学习
极端梯度提升模型
类别增强型提升模型
力学性能
hydrogenated nitrile butadiene rubber(HNBR)
machine learning
extreme gradient boosting(XGBoost)
categorical boosting(CatBoost)
mechanical property
