A Study of Spectral Envelope Method for Multi-Cause Diagnosis using Industrial Data

Petroleum refineries are complex systems that consist of multiple integrated units. This situation makes it difficult to track down the root cause of abnormal situations that occur during production. It is noted that abnormal situations usually trigger plant-wide oscillations in a number of measured process variables. Therefore, root cause detection is often attempted to be carried out by examining these trends in process data. Observing multiple effects and underlying problems at the same time presents a challenge in determining the root cause by examining trends only. In this study, spectral envelope is used to detect oscillations by identifying the variables and categorizing them based on a statistical hypothesis test which produces Oscillation Contribution Index (OCI) in order to isolate potential root cause variables. Two distinct abnormal events in the hydrocracker unit that occur simultaneously were successfully isolated and the root causes could be assigned by using the spectral envelope analysis.