聚丙烯酸酯类T_g的量子化学-神经网络研究  

Quantum Chemistry-ANN Methods Study on Tg of Polyacrylates

作　　者：刘万强[1] 王学业[1] 李新芳 龙清平[1] 文小红[1] 李建军[2] 

机构地区：[1]湘潭大学化学学院,湘潭411105 [2]江苏出入境检验检疫局,南京210005

出　　处：《物理化学学报》2005年第6期596-601,共6页Acta Physico-Chimica Sinica

基　　金：湖南省自然科学基金(02JJY2019);湖南省中青年科技基金(01JZY2099)资助项目~~

摘　　要：用密度泛函方法在6-31G(d)基组上优化了38种聚丙烯酸酯类的结构单元,得到了其单元的量子化学参数,探讨了这些参数与聚丙烯酸酯类玻璃化温度(Tg)的关系.计算表明,影响聚丙烯酸酯类Tg的主要因素有结构单元的侧链长度、侧链的分支数、最高占据轨道能级、极化率、偶极矩、等体积热容和热力学能等参数.用模式识别方法(偏最小二乘法)讨论了这些参数与Tg的定性关系,两类Tg大小不同的聚合物基本分布在不同区域,用逐步回归和人工神经网络方法建立了这些参数与Tg的定量关系,2种方法的预测结果与实验值的相关系数分别为0.9753、0.9985,标准偏差分别为18.42、4.25,预报结果与实验值基本一致.The mechanism and affecting factors of the glass transition for polymers have been analyzed. The structural units of thirty-eight polyacrylates have been optimized and their quantum chemical descriptors have been obtained by DFT/6-31G(d) method. The calculated results indicate that the length of side chain, number of side chains, polarizability, dipole moment, E-HOMO, heat capacity at constant volume, and thermal energy are the main factors affecting glass transition temperature(T-g). The regularity of T-g for polyacrylates are discussed by the pattern recognition method (PLS) with quantum chemical descriptors as features. The two classes of polymers with different T-g distribute in different regions. The quantitative relationship have been studied between these descriptors and T-g by stepwise regression and BP-ANN (back propagation artificial neural network) methods. The correlation coefficients between the predicted and experimental T-g for the two methods are 0.9753 and 0.9985, and the standard deviations are 18.42 and results 4.25, respectively.

关 键 词：丙烯酸酯类 Tg 人工神经网络方法 密度泛函方法 量子化学参数 偏最小二乘法 模式识别方法 结构单元 玻璃化温度 轨道能级 定性关系 热力学能 聚合物基 定量关系 逐步回归 相关系数 预测结果 预报结果 实验值 计算表 分支数 

分 类 号：O631.4[理学—高分子化学]