By bestowing our first-ever Plant Innovation Award, the staff at Chemical Processing are honoring significant improvements to existing operating facilities by the insightful engineers who run these plants on a daily basis. We received 13 nominations from companies in various sectors: chemicals, pharmaceuticals, petrochemicals, power generation and food processing.
We thank the five members of our editorial board who judged the entries: Vic Edwards, Aker Kvaerner, Houston; Tim Frank, The Dow Chemical Co., Midland, Mich.; Ben Paterson, Eli Lilly and Co., Indianapolis; Roy Sanders, PPG Industries, Lake Charles, La.; and Jon Worstell, Shell Chemical, Houston.
Our judges gave their highest rankings to:
• Jim Sturnfield, South Charleston, W.Va.-based senior specialist for The Dow Chemical Co., Eli Maldonado, Houston-based commercial manager, Dwight Pesek, process control technician, James Morgenroth, technical adviser, Rodney Hodde, production leader, and Seadrift Energy System Operations for implementing the G2 expert system from Gensym Corp., Burlington, Mass., in conjunction with Visual MESA, a closed-loop optimizer, from Nelson and Roseme Inc., Walnut Creek, Calif., at Dow’s Seadrift, Texas, site.
• Geovanna Nazario, principal engineer, Adalberto Maldonado, manufacturing manager, Diana Santiago, senior principal engineer, Marylin Roque, technical services engineer, Diana Rodriguez, documentation manager, and Carlos Santiago, lead instrument technician for Baxter Healthcare Corp., based in Guayama, Puerto Rico, for using model predictive control (MPC) from Emerson Process Management, St. Louis, in conjunction with statistical process control (SPC) to eliminate failed acetone batches by improving operation of an acetone recovery column. Francisco Feito, director, and Ruben García, director, sponsored this project (Figure 1).
Dow gets an energy boost
Last year, Dow Chemical saved $1.75 million at its Seadrift petrochemicals plant by reducing overall energy demand, including electricity and natural gas, due to the use of G2 and a closed-loop optimizer. Sturnfield estimates 2005 savings could reach $3 million resulting from increased natural gas prices, as well as other changes that have been made at the plant.
G2 is an object-oriented expert system software platform that captures operations expertise in the form of rules, procedures and models to infer production conditions and make supervisory control decisions. It reduces the amount of programming required to run the optimizer for the many possible operating conditions that exist at a site that produces a variety of products.
After trying several open-loop optimizers to improve operation of the cogeneration plant, a team led by Sturnfield decided to switch to a closed-loop optimizer, which would require less input from the operators. Visual MESA and G2 were installed in 1997. However, the optimizer was initially run in open-loop mode. As the system was debugged by Dow personnel, Sturnfield says they started closing loops, beginning with the heat-recovery steam generators (HRSG) and steam turbines (Figure 2).
The team’s efforts then turned toward implementing the system on cogen units at two other sites. The focus returned to the Seadrift site in 2002 when the scope of the project was expanded to include a new cogen unit. This necessitated additional training for the operators so they would trust the system and not turn it off when they felt the system’s recommendations were incorrect.
For example, the optimizer recommended running the turbines at partial load, but the operators were concerned that the hydrogen-rich fuel fed to the turbines might flash back and force a shutdown. The operators were encouraged to bring such problems to the attention of the engineers, who then used G2 to implement guidelines to the optimizer on the relationship between fuel richness and the limitation on partial loading of the gas turbine (Figure 3).
Once staff started seeing the value of the optimizer, use of the system increased; during a four-month period, the amount of on-stream time the system ran in closed-loop mode increased from less than 10% to more than 90%. During the past two years, the optimizer has run in closed-loop mode for more than 98% of the time.
“We are looking to use G2 to maintain the whole system,” Sturnfield says. “The degrees of freedom we have for operating the plant are affected by decisions that were made yesterday,” so he is analyzing the results from G2, which will help Dow implement it in other parts of the cogen plant, as well as at other sites. Sturnfield is also working toward integrating the optimizer and G2 with advanced control of the cogen unit.
Our judges were impressed by the immediate results the system provided. “This is a good example of what can be achieved through optimization of the overall work process for faster/better decision-making,” Frank says. Edwards adds, “This is [a] significant plant improvement through advanced process control.”
Another recognizes the value of optimizing energy resources. “While many chemical manufacturing facilities have cogen units, they are not operated as an integral part of the site, thus many of the possible gains are not realized,” Worstell says. “As we enter another period of world energy shortage, cogen can help ease the impact if its operation is optimized with respect to the operation of its parent facility.”
Baxter doesn’t stand still
Last year, engineers at Baxter improved operation of an acetone recovery column (the Xc still) through use of MPC. The throughput and quality of acetone increased and failed batches were eliminated, thereby making it unnecessary for Baxter engineers to spend time documenting and investigating off-spec material.
The Xc still recovers acetone from the effluent of an upstream column for reuse (Figure 4). The recovered acetone must meet a specification of less than 3 wt-% water. During 2003, 615 batches of acetone were processed, 18 of which failed the water specification, leading to significant delays pending the investigation and documentation of each failed batch.
The team used SPC to analyze data from 140 batches, which showed that the acetone had an average water content of 2.3 wt-%, a Cp of 1.24 and Cpk of 0.54. A low Cp indicates a high degree of spread in the data (a Cp of 2 or greater corresponds to Six Sigma performance), whereas the higher the Cpk, the closer the data are to the target. Hence, the data from these batches indicated a lack of control and significant room for improvement.
The first step toward improving column operation was to reconfigure the control scheme and improve the tuning so it could run in automatic mode. Despite the fact that six batches exceeded the 3 wt-% water specification, data from 125 batches showed a 35% reduction in average water content to 1.5 wt-%. Although the average water content went down (Cpk = 0.86), the spread in the data increased (Cp = 0.88).
Other columns onsite were controlled using MPC, so the team decided to implement it on the Xc still. MPC is an add-on module available from Emerson for the DeltaV distributed control system (DCS). The module enables concurrent control of multiple process constraints, rather than managing them as individual loops or variables. One block can monitor up to four different variables and anticipate the expected behavior, thereby applying several corrective outputs as necessary to maintain the optimal column performance. Feito says they use one control block to monitor four inputs and two outputs on the acetone column.
Once MPC was employed, the average water content of 31 consecutive batches was reduced to 1.2 wt-% -- none of them failed – and SPC showed a Cp of 8.98 and Cpk of 7.39. Santiago says there have been no failures since MPC was implemented -- about 600 consecutive batches to-date have been within specification.
Maldonado says that when batches failed the capacity to reprocess the offspec acetone was limited, which might necessitate discarding material and mading up with fresh acetone.
Our judges were impressed by the Baxter team’s innovative use of MPC. “This technology is used widely in the petrochemical industry, but not that much in pharmaceuticals,” Paterson says. Frank adds, “This is a good example of what can be achieved through implementation of improved process control schemes and operating discipline.”
Another recognizes the value of making improvements without a hefty capital investment. “Most batch chemical manufacturing facilities are underutilized,” Worstell says. “Employing computer scheduling software at such facilities provides a means for achieving capacity expansions without capital investment. Implementing such software will increase production and reduce costs, thereby increasing the standard of living for the world community.”