At the 3M plant in Cottage Grove, Minnesota, recurring problems had prompted the materials resource division to plan a distributed control system (DCS) migration. “We had a daunting challenge,” explained Matt Rauschke, senior resident engineer at 3M, who led a presentation at this week’s Emerson Global Users Exchange. “How do you connect the tank farm’s three different control systems?”
The first six months of the project were spent migrating two of the systems from PROVOX to DeltaV, while the third system remained a Modicon programmable logic controller (PLC-based) system. But the longer-range impact of the upgrade was destined to affect three main areas—environmental, health, safety & regulatory (EHSR); operations; and accounting/bookkeeping.
“We needed to migrate the DCS for operational improvements,” explained Rauschke. “But there were specific issues to address. Spills were a problem, so we needed to improve our EHS&R risks and consequences. We had two accounting systems, and they weren’t reconciled.”
The DCS migration needed to happen because parts were becoming difficult to find for the PROVOX system. “System reliability was extremely poor,” admitted Rauschke. “The pump time altered frequently, and a building supervisor would have to come over and reset the pump. Plus, we didn’t have a standardized solvent charging program.”
The biggest problem with the leak detection system was it wasn’t a leak detection system at all. “It relied on a set maximum pump run time,” explained Rauschke. “We had the potential for a spill the entire length of the pipe.”
Accounting was inaccurate. “We had 6% of our solvent that wasn’t accounted for properly,” said Rauschke. “We use a lot of solvent, so that was a big problem. We had no way to reconcile true usage with run cards.”
The tank farm had a lot of problems to address. “Early on in the migration project, we decided to move away from the pressure-based system and move to an on-demand system,” explained Rauschke. “The new system would require an on signal from each user. The biggest challenge was to make it easy for the operators. We tried to tailor the system to make it invisible and intuitive to the operators. The operator would open a valve on the manifold, and the system would turn on the pump and then close the automated valve. We also installed a pushbutton system with limit switches to select which solvent they wanted. And we improved our wireless coverage in the tank farm. This was critical later in the project.”
For the leak detection system, a new software-based system was designed to reduce the potential of a 7,000-lb spill. “The new system would use a flow rate, and it allowed us to get down to 250 lb in a spill event,” said Rauschke.
And, for tanker unloading, the old system only made sure the tanks wouldn’t overfill. “In the new system, the operator would open a limit switch, and the HMI would require they look at tank,” explained Rauschke. “Rail cars would be weighed as they came in and then weighed as they left. That would tell us how much we actually used in our plant.”
As the project progressed, the two databases needed to be compared. “As you get information out of DeltaV, you have to go through the historian,” explained Kyle Nystrom, operational intelligence engineer for Novaspect, an Emerson local business partner in the Midwestern U.S. “We decided to put everything into the event journal. We ran into a few issues, in terms of syntax, so we had to create a relational database and use a third-party software called BridgeWorX.”
At 3M Cottage Grove, there are three divisions, explained Colin Singer, project manager, also with Novaspect. “Specialty chemical is the largest solvent user,” he said. The chemical pilot plant was the least technically involved implementation. And the adhesives pilot plant is the Modicon system.”
While the Modicon system wasn’t migrated, the other two were upgraded to the DeltaV solution, which included a three-level architecture—solvent distribution locations, coordination modules and solvent supply. “Each location or requester would communicate with the coordination module,” explained Singer. “If it were a new request, it would contact the solvent pump to start. There was a coordination module for every solvent that was used. Before our upgrade, there was no level of automation at the drum filling station. We put limit switches on all of the valves.”
Lessons learned the hard way
With any project startup there are issues with commissioning. “We discovered some shortcomings,” said Rauschke. “At first, the pumps were tripping constantly, and recycle lines didn’t work as intended. We were also struggling to get feedback from operators for what was tripping the pump.”
There was also a fundamental problem in how the system was designed with a pump and variable frequency drive for flow control. Leaks were to be detected by comparing the flow rate predicted by the VFD and pump curve against a Micro Motion Coriolis meter on the other end of the solvent movement system. “But the VFD could not match the performance of the Micro Motion flow meter,” said Rauschke. “Startup dynamics and drift were the two major issues.”
Seasonal variations were especially troublesome for leak detection. “We had to abandon the leak detection system for a while because of the temperature variation in Minnesota,” explained Rauschke. “The solution was to install wireless conditioned differential-pressure (DP) flow transmitters and temperature transmitters. There were cost savings of more than $150,000 over hardwired transmitters.
The tank farm control system migration and on-demand pump system costs totaled about $600,000. “It’s a much more reliable safety system now,” assured Rauschke. “The solvent reporting discrepancies decreased from 6% to 1.3%, and we reduced maintenance costs by reducing pump cycling. But we can’t know it’s working until it doesn’t work, and we hope we never see that.”