2025 Spring Meeting and 21st Global Congress on Process Safety

(42a) Improving Ethylene Furnace Operations with Infrared and Machine Learning Technologies

Authors

Yi Liu - Presenter, Westlake Digital
Michael Dessauer, Dow Chemical
Tube metal temperatures (TMT) of radiant coils are one of the critical limiting factors for the ethylene furnace run-length. TMT has traditionally been measured using pyrometer surveys, which benefit from operator experience but can be limited by subjective observation skills and the difficulty of accurately targeting the hottest tube. One significant drawback of using pyrometers is that they expose operators to extreme heat sources for prolonged periods, which can pose serious safety risks. Furthermore, due to multiple operators across several shifts shooting TMTs, fairly significant error is introduced along with consistent and repeatable readings being recorded. This introduces significant error in the available TMT data, which is utilized for tactical and strategic operational decision-making.

Infrared (IR) technologies with cooled sensors offer temperature measurements with high repeatability, resolution, and sensitivity. Combined with machine learning (ML) technologies for object identification, IR technologies present a superior alternative for monitoring TMT and providing enhanced operational data that can be overlayed with inspection and maintenance data. Infrared image analysis can efficiently provide temperature information on IR images, where we employed image segmentation technology for object identification.

We have conducted trials of IR TMT surveys over two years at the Westlake Ethylene facilities. The IR surveys have become a supplementary TMT measurement tool for the operation team and maintenance team. This paper discusses various applications of the TMT surveys, including TMT tracking, hot spot identification, and heat balance improvement for the furnace. Additionally, we aim to explore the most cost-effective implementation of the IR inspection operation procedure in the future.