摘要
在对传统的数据挖掘技术加以改进的基础上,利用均匀实验设计、灰关联分析、逐步回归变量筛选、无阀值逐步回归和非线性偏最小二乘法等多种数学方法,对软质聚氯乙烯阻燃配方体系作研究实验。证明灰关联排序和灰色优势分析,对配方设计与分析行之有效,改进后的建模方法可以在一定程度上,避免小样本带来的拟合误差。通过对配方体系5项指标(氧指数、烟密度、拉伸强度、延伸率和热释放速率)的数学模型分析,深入探讨体系中,添加剂之间的作用及其对体系各种性能的影响。
The flexible PVC Formulation System is studied by data mining technology, such as homogenous experimental design, gray correlation analysis, step regression, partial least squares regression etc. Grey correlation sorting and gray preponderance analysis is proved to be available. To avoid fitting error of few samples the author make a suggestion of step regression without threshold and simple non-linear partial least squares regression. Mathematic models of five specifications (limited oxygen index, smoke density, tensile strength, elongation and heat releasing rate) are analyzed to probe into the formulation system.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2007年第6期750-753,共4页
Computers and Applied Chemistry
基金
北京市青年科技骨干培养基金资助项目
关键词
聚氯乙烯
阻燃材料
数据挖掘
数据建模
配方研究
PVC, flame retarded material, data mining, data modeling, formulation research