Cross Oil Refinery, a 7,500 barrel-per-day lube oil refinery founded in 1930 in Smackover, Arkansas, produces products ranging from diesel fuel to asphalt and specializes in high-quality transformer oil. “We haul our own oil, refine it, package it and ship it ourselves,” said Christopher Brantley, E&I programmer for Martin Operating Partnership which operates the refinery. “Running a variety of viscosities and properties on the same process equipment, transmitters and controllers means we have to change the process for different products, each with its own set of tuning parameters.”
Brantley co-presented “Control Improvements on a Mild Hydrotreater” with James Beall, principal process control consultant, Emerson Automation Solutions, this week at the Emerson Global Users Exchange in San Antonio.
Ten years ago, the refinery was running on single-loop controllers. Now it has a DeltaV system with more than 1,000 tags; Allen-Bradley PLCs; a mixture of Fisher, Flowserve and Cashco control valves and actuators; basic Foxboro, Rosemount and Yokogawa transmitters; and is increasingly standardizing on Rosemount radar level transmitters and Micro Motion Coriolis flowmeters.
“Emerson called me about advanced process control (APC) and got me to listen when they said we can change tuning parameters on the fly,” Brantley said. “We had just gotten our operators used to PID faceplates so we took our time, got quotes, worked together and came up with a pilot project to tune a few loops on the mild hydrotreater (MHT).” The objectives were to improve MHT performance at minimal cost and with good ROI, including training for the control system staff.
Loop tuning stumbles on nonlinearities
“The staff at Cross Oil were not used to working with consultants and the operating culture is very conservative—if there was any risk to the process, they’d rather not try to improve,” said Beall. “We had two days to show how APC could reduce variability, so we installed the [Emerson] EnTech loop analysis toolkit and went to work. We focused on level loops because they have the most potential to minimize flow variability.”
While trying to tune the loops, they found many problematic instruments and valves, which caused a lot of cycling in the process and made it hard to quantify improvements. “We found we could make a little improvement by tuning, but big results by getting the process fixed,” Beall said.
“I had been asked to tune these loops before, and I could get a level loop absolutely flat,” Brantley said. “But that caused problems in the rest of the refinery because the flow valves were moving too much. We needed to understand how much level could vary so the flows could be smooth.”
The team did step tests to determine lambda—the arrest time—and found the response would not track the controller. Step tests were done in cooperation with operations to be sure the process would not be disturbed.
Often, the valves would not respond correctly. In one case, the level gauge float was incorrect for the process specific gravity, and the valve also had 4-6% deadband. They left the loop improved, but still cycling. “When the valve was replaced, the flow smoothed out, and so did the level,” Beall said. “Tuning the loops is the easy part, the hard part is eliminating the nonlinearities. Instruments and valves are like my cows, it seems like they’re always trying to commit suicide.”
The two-day project tuned seven loops, and when the nonlinearities introduced by valve and transmitter problems were resolved, “stabilized the process and gave much tighter control, which allowed operators to shift their focus toward distilling higher-quality products and recovering more of the crude’s higher-profit yields,” Beall said.
Virtually instant payback
The staff gained a much deeper understanding of control loops, and how to analyze a group of loops and manipulate them to better suit overall process goals. Brantley said, “We also gained a deeper knowledge of DeltaV PID modules, and incorporated several parameters in the PID module that we’d never used and had not even thought about.
“I learned a lot about tuning. There’s nothing like having an experienced person working with you on your own process.”
The two-day project is saving $245,000 per year, which calculates to a payback of 10 days. “We had no trouble getting approval for Jim to come back, this time for two weeks,” Brantley said. “This time, we implemented some feedforward along with regulatory control, and the payback was one month.
“I’m looking forward to doing some model predictive control.”