Automatic Fusion Algorithm Based on SDG for Fault Diagnosis of Petrochemical Process

Abstract

In petrochemical process, the kernel task of avoiding abnormal situation is fault diagnosis of the process. As signed directed graph (SDG) can reflect the path of fault propagation clearly, it is a hot spot in fault diagnosis of petrochemical process currently. However, basic SDG is poor in resolution and insensitive to early fault, so it is needed to introduce other algorithms to solve the shortcomings. It proposed an automatic fusion algorithm based on SDG which including fuzzy algorithm and principal component analysis (PCA) in this paper. It applied principal component analysis method to detect the presence of faults, and identified the possible failures of the nodes at first, then reasoned root cause by SDG combining with fuzzy algorithm. The simulation experiments on a distillation system shows that this automatic fusion algorithm improve the reasoning speed and fault resolution greatly.