摘要
符号有向图(SDG)是用来表示大规模复杂系统中变量之间因果影响关系的一种重要工具,但其存在一些不易克服的缺点.为此,首先提出一种新的模型——概率SDG模型,用条件概率描述故障之间的传递关系;然后在概率SDG模型的框架下,提出一种故障分析诊断的推理方法,即利用图消去算法和连接树算法进行贝叶斯推理,并计算出故障概率.最后以65 t/h锅炉系统为例,研究建立其概率SDG模型,并在此基础上验证了上述模型和推理方法的有效性.
Signed directed graph （SDG） model is a significant tool to express cause-effect relationships between variables in large-scale complex systems, but it also has some disadvantages that are difficult to overcome. A new kind of model, probabilistic SDG model, is proposed to describe the dependence relationships with conditional probabilities. In the framework of probabilistic SDG model, inference approach is presented, which implements Bayesian inference with elimination algorithm and junction tree algorithm to calculate the fault probability. Finally, a probabilistic SDG model is established for a typical instance of 65 t/h boiler system, which proves the validation of the model and inference approach.
出处
《控制与决策》
EI
CSCD
北大核心
2006年第5期487-491,496,共6页
Control and Decision
基金
国家863计划项目(2003AA412310)
关键词
符号有向图
安全评价
故障分析
贝叶斯网络
Signed directed graph
Hazard assessment
Fault analysis
Bayesian network
