摘要
我们将随机取样高维模型映射(HDMR)方法应用于RRKM/ME计算的全局不确定性分析之中~[1],得到如下结论:1.二阶及以上的灵敏性系数对于RRKM/ME方法得到的速率常数的不确定性影响很小;2.当把输入参数按照能量参数、碰撞参数和频率参数分为三类以后,每一类参数引起的速率常数不确定性会随着温度和压力单调变化。基于以上结论,本工作中我们在之前研究的基础上进行了更为细致的计算,力求得到上述参数随着温度和压力变化的定量描述,最终得到一个能够广泛适用的不确定性评估规则。如考虑由碰撞参数引起的不确定性随温度和压力的变化关系,如果能够得到高压极限和低压极限下的不确定性随温度和压力的变化关系,我们也许能够推导出任意温度和压力下对应的局部不确定性。值得一提的是,本工作通过运用人工神经网络(ANN)算法~[2]极大地降低了不确定性分析中需要的计算量。
The random sampling high dimensional model representation(RS-HDMR) method has been used to perform global uncertainty analysis for RRKM/Master Equation simulation.~[1] Two conclusions have been made: 1. The second and higher order sensitivity is unimportant for uncertainty of RRKM/ME rate constants; 2. If the input parameters were divided into energy parameters, collisional parameters and frequency parameters, the partial uncertainties of computed rate constants monotonically increase or decrease with temperature and pressure. Based on these conclusions, in the present work, we have extended the previous study to perform more rigorous uncertainty analysis on RRKM/ME rate constants aiming to quantitatively evaluate general rules for the temperature and pressure dependence of the uncertainties for computed rate constants. We have found that it might be possible to deduce partial uncertainty at any temperature and pressure if one knows the temperature and pressure dependent uncertainty of k at both high and low pressure limit. It should be noted that an artificial neutral network(ANN) algorithm~[2], which largely reduced the computational cost of uncertainty analysis, was used in this work.
引文
[1]Xing,L.-L.;Li,S.;Wang,Z.-H.;Yang,B.;Klippenstein,S.J.,Zhang,F.,Combust.Flame 2015,162,3427-3436.
[2]Li,S.;Yang,B.;Qi,F.,Combust.Flame 2016(In Press)