应用变量因果序分析的符号有向图建模方法

New Methodology of Signed Digraph Modeling Based on Causal Ordering Analysis

针对传统符号有向图(SDG)建模方法容易引入虚假因果关系的问题,利用因果序分析对系统各阶自含结构进行逐层分解,通过各阶变量完全子集迭代解析平衡结构变量的层级因果关系,以及典范式积分链的约简导出动态结构变量的层级因果关系,运用中间变量的衔接作用,结合变量系数辨识定性的因果影响效果,实现了平衡、动态及混合结构的线性系统SDG模型的构建.以串联存储系统为实例进行验证,结果表明该方法消除了以系统方程等式关系直接导出SDG模型引入的虚假因果关系,且辨识出多处潜在的因果关系,与传统的SDG建模方法相比,以该方法构建的SDG模型精度更高,而且完备性较好.

Focusing on signed digraph（SDG）models holding inaccurate information to induce spurious solution,causal ordering analysis is introduced to resolve causal dependent relation of system variables.The hierarchical causal dependent relation among variables in system equations is revealed by complete subsets iteration of self-contained structure of equilibrium structure and integral chain reduction in canonical form of dynamic structure.The causal dependent effect among variables is obtained by coefficients identification.SDG models are then constructed for linear systems of three forms of equilibrium structure,dynamic structure and mixed structure based on causal ordering analysis.The experimental instance indicates that in the SDG model constructed the spurious information is removed and the accuracy of SDG model is improved effectively.