Cause analysis of representative troubles at distillation tower using discriminant analysis

In order to improve crude distillation unit (CDU) operating efficiency, a cause analysis method to identify CDU anomalies based on discriminant analysis was proposed. The method identifies sensors in CDU that behave different between during anomaly and normal period using discriminant analysis. To detect cause of troubles with the identified sensors, category table to categorize sensors based on domain knowledge and statistical analysis using histogram were introduced. Binary distillation experiments using a mixture of ethanol and water carried out on test equipment at Universiti Teknologi PETRONAS (UTP) confirmed that the method could detect causes in 6 troubles out of total 7 heat exchanger related troubles that were experimented in the equipment. © 2017 IEEE.