(155d) Using thermal imaging and a cloud platform to optimize cracking furnace efficiency in real time

Using kinetic models to predict the behavior and performance of cracking furnaces is state of the art. However, in many cases these predictions do not match reality in the plant resulting from problems with the measurements (e.g. coil outlet temperature) or the tuning of the model. Neither the predicted values from the model nor the manual tube metal temperature (TMT) measurements are really precise.

Using infrared cameras with AI based image recognition software we can measure the actual temperature profile along the coils in real-time 24/7 during cracking and decoking. By continuously comparing the TMT profile predicted by the kinetic model with the TMT measured by the camera we can:

• Prevent coil overheating resulting from flame impingement or hot spots resulting from flow restrictions
• Extend furnace run-length closer to material limits
• Shorten decoking times through temperature controlled decoking
• Detect measurement and model errors and correct them so that the kinetic model output actually matches reality.

For a long time the ability of furnace experts from licensors to support furnace operators in real-time was limited by the possibility to share real-time process data. Modern cloud platforms like Microsoft Azure are providing the infrastructure for a collaboration of licensors and operators that meets the stringent IT security requirements for operators of critical infrastructure.

This paper will present Linde’s solution combining their real-time kinetic model with continuous thermal imaging of the coils accessible through a cloud platform, and the first experiences with their implementation for operators of steam crackers in Europe, Middle East, Russia and Asia.