Digital Twin for Takamul Smart Technologies
Takamul Smart Technologies is a Saudi Arabian IT integrator and is developing a smart maintenance platform. They are currently piloting this platform for a big gas processing plant. The control of such a processing plants can be quite tricky, with many control valves (and flares) and many elements influencing the process. In order to make the control easier and see potential issues coming sooner, we developed two digital twins for the plant.
At first we focused on the main gas compressor. Industrial gas compressors have operating characteristics and safety systems in-place which shut down the compressor in case of unsafe conditions, such as an almost surging or stalling compressor. Historical data of various vibration, pressure, flow and temperature sensors of the compressor was available of many years. We used this to make predict pressure deviations, one of the main indicators for an upcoming surge or stall, 30 minutes ahead of time.
Secondly, we pointed our Automated Machine Learning towards a multitude of cooling towers. Gas is pumped through these towers to decrease the temperature after compression. The pumps are driven by a small turbine, and the temperatures of those turbines can reach critical levels where wear & tear speed increases significantly. It is therefore important to ensure operating temperatures below these critical levels by for example adjusting the mass flow through the tower. From the same database, we were able to collect many years of operating data. We developed various models to predict temperatures and vibrations in the pump & turbine 20 minutes ahead of time.
Together with Takamul, we integrated our machine learning models directly into the platform. The process engineers are stoked about the predictions as it allows them more time to act instead of constantly react.