论文部分内容阅读
Q-e方程在解释自由基共聚合中单体的活性时相当有效.为得到可靠的Q,e活性参数结构-性能定量关系(QSPR)模型,采用密度泛函理论(DFT)UB3LYP方法在6-31G(d)基组水平上对60种结构为CH3C1H2—C2HR3·自由基(C1H2=C2HR3+CH3·→CH3C1H2—C2HR3·)进行了计算.两组包含原子电荷及前线分子轨道能级的量子化学参数分别用来建立Q,e活性参数的人工神经网络(ANN)模型.通过试差法调整网络参数得到最佳ANN模型,两者均为3-5-1结构.参数Q,e预测值与实验值接近,测试集相关系数分别为0.990(rms=0.269)和0.943(rms=0.331).而且Q,e模型的外部验证系数qe2xt分别为0.980和0.873,这结果显示两模型具有好的推广预测能力.因此本工作介绍的ANN模型是精确而可靠的;从自由基CH3C1H2—C2HR3·结构获得参数预测Q及e值是可行的.
Qe equation is quite effective in explaining the monomer activity in free radical copolymerization.In order to get a reliable quantitative QSPR model of Q, e activity parameters, the density functional theory (DFT) UB3LYP method was used to determine the activity of 6-31G (d) Calculations of 60 structures of CH3C1H2-C2HR3 · radicals (C1H2 = C2HR3 + CH3 · → CH3C1H2-C2HR3 ·) at the base group level. Two sets of quantum chemical parameters, including atomic charge and frontier molecular orbital energy levels Respectively, to establish Q, e activity parameters of the artificial neural network (ANN) model.Adopting trial and error method to adjust the network parameters to get the best ANN model, both of which are 3-5-1 structure.Parameters Q, e predict value and experiment (Rms = 0.269) and 0.943 (rms = 0.331), respectively, and the external validation coefficients qe2xt of Q, e model are 0.980 and 0.873, respectively. The results show that the two models have good ability to predict and predict Therefore, the ANN model introduced in this work is accurate and reliable. It is feasible to predict the Q and e values from the parameters of free radical CH3C1H2-C2HR3 · structure.