On the application of betweenness centrality in chemical network analysis: Computational diagnostics and model reduction
On the application of betweenness centrality in chemical network analysis: Computational diagnostics and model reduction
复制标题
介数中心性在化学网络分析中的应用：计算诊断和模型简化
DOI：
10.1016/j.combustflame.2015.05.011
复制
发表时间：
2015
影响因子：
4.4
通讯作者：
C. Law
中科院分区：
文献类型：
--
作者：
Peng Zhao;Samuel M. Nackman;C. Law
在这项工作中，最短路径和介数中心性（BC）的概念被引入到燃烧系统中，其应用表明在化学网络分析和骨架机制生成。在针对给定的压力、温度和物质浓度建立化学网络的图形表示之后，定义度量BC以对由源节点和目标节点对之间的其他节点通过的最短路径进行排名，并且因此捕获网络中的任何节点对之间的主导间接动力学路径以用于计算诊断。因此，通过收集具有较大BC值的物种来保留控制途径。这样的概念，然后表明，物种的重要性指数，然后可以分配基于其BC值，以进一步指导骨骼机制生成。与现有方法不同，介数中心性方法考虑了物种之间的通量及其在化学网络中的相对位置。为了证明其潜在的效用燃烧研究，该方法被应用到GRI-3.0，LLNL和USC-Mech II机制，以确定在每个局部反应状态的化学网络中的重要途径，并从自燃和完全搅拌反应器（PSR）模拟中的所有反应状态样本开发骨架机制。BC排名的性能与有向关系图（DRG）、带误差传播的DRG（DRGEP）和敏感性分析（SA）方法进行了比较，并被证明在产生具有良好准确性和灵活性的骨架机制方面具有足够的实用性对于所研究的案例。
In this work, the concept of shortest path and betweenness centrality (BC) is introduced in combustion systems, with its application demonstrated in chemical network analysis and skeletal mechanism generation. After establishing the graphical representation of a chemical network for given pressure, temperature, and species concentrations, a metric BC is defined to rank the shortest paths passed by other nodes between the pair of source and target node, and as such captures the dominant indirect kinetic pathways between any pair of nodes in the network for computational diagnostics. Consequently, the controlling pathways are retained by collecting the species with larger BC values. Such a notion then indicates that the index of importance of species could then be assigned based on their BC values to further guide skeletal mechanism generation. Unlike existing methods, the betweenness centrality approach takes into account of both the fluxes between species and their relative positioning within the chemical network. To demonstrate its potential utility to combustion studies, the approach was applied to the GRI-3.0, LLNL and USC-Mech II mechanisms to identify the important pathways in the chemical network at each local reaction state, and develop skeletal mechanisms from all reaction state samples in auto-ignition and perfectly stirred reactor (PSR) simulations. The performance of the BC ranking is compared to the methods of directed relation graph (DRG), DRG with error propagations (DRGEP) and sensitivity analysis (SA), and is shown to possess sufficient utility in producing skeletal mechanisms with good accuracy and flexibility for the cases studied.