基于机器学习的CF残留长度预测和影响因素特征重要性分析OA
Machine Learning-based prediction of carbon fiber residual length and feature importance analysis of influencing factors
以CF/PA6复合材料为研究对象,分析了以密炼机制备CFRP时,转子结构、转速、螺棱间隙和PA6基体黏度对CF残留长度的影响,并基于机器学习算法对各因素的特征权重和重要性进行了预测和评价.结果表明,线性回归算法能够对CF残留长度对数平均值进行有效预测,而非线性算法中的决策树和梯度提升回归模型则能更精准地评价各因素的特征重要性.熔体黏度对CF平均残留长度的影响程度最高,特征重要性超过0.4;其次是转速,接近0.32;螺棱数量为近0.25;而螺棱间隙对CF残留长度的影响极小.
This study investigated the preparation of PA6/carbon fiber(CF)composites in an internal mixer,focusing on the effects of rotor structure,rotational speed,rotor flight clearance,and PA6 melt viscosity on the residual length of car-bon fibers.Machine learning algorithms were employed to predict and evaluate the feature weights and importance of these influencing factors.The results indicate that the linear regression model effectively predicts the logarithmic mean of CF residual length,while nonlinear models,specifically decision tree and gradient boosting regression,provide more ac-curate assessments of feature importance.Among the factors analyzed,melt viscosity exhibits the greatest influence on the average CF residual length,with a feature importance exceeding 0.4;rotational speed follows with a value close to 0.32;the number of screw threads contributes approximately 0.25;and rotor flight clearance shows the least impact on CF residual length.
杨文明;谢林生;周胜荣;杨伟光;徐成龙
华东理工大学机械与动力工程学院,上海 200237华东理工大学机械与动力工程学院,上海 200237华东理工大学机械与动力工程学院,上海 200237华东理工大学机械与动力工程学院,上海 200237华东理工大学机械与动力工程学院,上海 200237
化学化工
碳纤维增强复合材料纤维断裂纤维长度机器学习
carbon fiber reinforced polymerfracture of fiberlength of fibermachine learning
《中国塑料》 2026 (5)
42-47,6
评论