Fault Propagation Analysis and Causal Fault Diagnosis for Petrochemical Plant Based on Granger Causality Test

Abstract: 

Petrochemical plants have becoming increasingly large and automatic. There exist strong interdependencies between various units. Once any unit fails, it often triggers cascade failures as a chain reaction, resulting in significant production losses and catastrophic accidents. Fault diagnosis methodologies can unveil early deviations in the fault causal chain. Two main issues exist that are how to describe the interdependency in such complex system, and how to discover the root causes of the current abnormal event to support maintenance. Meanwhile in order to reduce the fault impact to the plants, quickly diagnosis of the root cause of the fault is also quite necessary.

In this paper, aiming to solve above issues, granger causality test is introduced to study the fault interdependency by analyzing the relationship between process parameters of petrochemical units and establishing an effect diagram of the process parameters.

When alarm occurs on condition monitoring system, the effect relationship diagram of the process parameters is used to elect related process parameters which haven’t exceed the alarming threshold but may indicate an incipient fault. Then the granger causality test is used on the selected parameters to do the test pairwise. According to the degree of the causal relationship of the process parameters, the fault quantitative cause and effect diagram can be established. By using the quantitative cause and effect diagram, the path with the biggest quantitative value of causal relationship can be considered as the most probable fault propagation path in the petrochemical units according to the current alarm. In this way the root cause of the alarm can be revealed easily.

The pilot application for FCCU and atmospheric and vacuum distillation unit in the case studies validates the effectiveness of the proposed method and its application value in the petrochemical industry.