Problem: Two historical problems found in AGRU of gas separation plant of a large gas separation operator in Thailand : i) significant changes of natural gas composition due to upstream source gas process upsets and ii) stripper reboiler tube leakage which causes high pressured hot oil to contaminate with low pressured amine solution in shell of reboiler. These two anomalous events could result in emergency shutdown due to control issues and significant decrease of acid gas removal system efficiency. They want to be able to detect and classify these two faulty cases as early as possible.

Benefit: We could predict around 7 hours prior to the plant upset due to natural gas composition variation, and around 27 hours prior to the shutdown due to the tube leak problem. This early detection would give enough tie for the operator to provide appropriate action and prevent the serious plant upsets, or even the shutdown.