基于因果拓扑图的工业过程故障诊断

基于因果拓扑图的工业过程故障诊断方法,将过程知识与数据驱动故障诊断方法结合,有效解决了故障定位和故障传播路径辨识问题。在因果拓扑图的基础上,基于偏相关系数提出一种相关性指标(correlation index,CI)定量衡量因果拓扑中变量间的相关性,实现变量间因果性和相关性的良好结合。为得到准确的故障检测结果,采用概率主元分析(PPCA)对CI指标进行监测。在检测出故障后,应用重构贡献图(reconstruction-based contribution,RBC)和因果拓扑图,并引入加权平均值的概念辨识出最可能的故障传播路径。将提出的方法用于带钢热连轧过程,结果表明,基于因果拓扑图的故障诊断方法能够准确地定位故障源,辨识故障传播路径。

Fault diagnosis for industrial processes based on causal topological graph

With the combination of the process knowledge and data driven methods, the fault diagnosis method based on causal topological graph could effectively deal with the fault location and fault propagation identification. A correlation index(CI)based on partial correlation coefficient was applied to the causal topological graph to analyze the correlation between variables quantitatively. The proposed CI was monitored via probabilistic principal component analysis method(PPCA)for fault detection. The concept of mean weighted value and causal topological graph were introduced in order to identify the optimal fault propagation path based on reconstruction-based contribution(RBC)after detecting a fault. The effectiveness of the method was verified by the application of hot strip mill process(HSMP). The results showed that the proposed method could effectively identify the fault roots and propagation paths.