March 22, 2021: Everything happens for a reason. At least within the confines of the refinery and petrochemical plant piping and vessels, this is true. But that reason often isn’t obvious and takes the right combination of skills and tools to reveal it. That’s one of the reasons I was attracted to the operations side of the process industry. Armed with those beautiful laws of thermodynamics, I had an opportunity to play detective on a regular basis.
I recall one particular case involving repeated near misses on exceeding emissions limits from a sulfur plant. At random intervals, the Tail Gas pressure control valve would jump open, releasing Tail Gas to an incinerator and bumping up against the permitted SO2 emissions limits. At the beginning of the investigations, the list of potential causes was practically endless. Was it a problem in the Tail Gas Recycle Compressor? Maybe there was a plug in a Condenser Seal Leg? Was the pressure spike even real at all, or was the pressure control valve responding to a phantom instrument anomaly?
Starting at the scene of the crime and painstakingly checking every instrument and detail back to the source, after 4 hours I discovered a faulty connection on a temperature transmitter used in the temperature-pressure compensation for the oxygen flow meter feeding the Claus Furnace. This caused a spike in oxygen flow, upsetting the stoichiometric ratio for the catalyst beds downstream, leading to poor conversion and overwhelming the Tail Gas Recycle Compressor.
What took hours in 1997 to assess can now be done within minutes with today’s sophisticated digital tools. Using advanced data analytics, continuous instrument diagnostics can be applied to every instrument in the field, which also provides immediate indication of a failure. With a hybrid 1st principles-statistical model, the process can be accurately modeled to highlight the source of process upsets. This model continuously learns the nuances of one particular plant to better forecast the response to changes in feed, operating conditions, or other disturbances. Even better, this model creates so-called “soft sensors,” revealing transport property data about the process, where no transmitter in the field exists.
So, whether it’s revealing insights not previously available with traditional tools, or simply taking the data analysis burden off the shoulders of the plant personnel, digitalization can increase the effectiveness and expand the capabilities of today’s operators, engineers, and plant managers.