Sweeter Simulation

"I love it when a plan comes together," says Col. John "Hannibal" Smith. The cigar-chomping leader of "The A-Team" was talking about his group's cartoonish, live-action TV adventures in the 1980s. However, he might as well have been discussing simulation's multiple and merging roles in many process applications. The confidence and satisfaction is the same.

Slow and costly models applied to only big-ticket applications have long since been joined by faster and less expensive simulations in a wider variety of settings. However, constantly improving and higher fidelity models, new variables and parameters, better software, more powerful computers, 3D displays and other advances are also blurring the lines between simulation's usual categories. 

Most notably, static simulations used for design and configuration are being linked to dynamic simulations for operations and training, and these have been enhanced by closer-to-real-time data, which allows them to optimize actual operations, performance and products. As a result, tying and unifying simulation into one multi-functional bundle is letting users pick the capabilities they need without having to implement several different solutions.

For example, Exxon Mobil estimates its process applications worldwide are saving about $750 million per year by using ROMeo optimization software, Pro/II simulation software and other solutions from Invensys Operations Management, according to Joe McMullen, product manager for SimSci-Esscor at Invensys. ROMeo secures and reconciles measurements from Exxon's components, simulates subsequent conditions and courses of action, and recommends which ones will optimize the process and make it most profitable. Likewise, he adds that Royal Dutch Shell is also rolling out ROMeo in several refineries, and so far gains about $1000 in optimization benefits for every $25 it spends on support.

"The ultimate goal is to use one calculation engine for many different purposes," says McMullen. "As long as training simulators and process control systems are on different platforms, keeping them synchronized will continue to be an issue. By contrast, our dynamic modeling tool lets users easily create a model for use in a training simulator by just pressing a button, and our EyeSim 3D, virtual reality training simulator is starting to gain traction too."

Propylene Process Makes Its Own Model

Similarly, the biggest oil company in Sweden, Preem, has integrated Hysys simulation models from Aspen Technology into its DMCplus controller to develop new advanced process controls (APCs) and cut product variation in half. Preem has two refineries including its Preemraff Lysekil refinery that processes 11.4 million tons of crude oil per year (Figure 1), and 500 retail gas stations. In fact, Preemraff was where the company recently addressed the challenge of controlling its propylene/propane (PP) splitter, which is used to separate C3 streams into 99.5% pure, chemical-grade propylene and 98.5% pure propane. Any deviations in quality could affect Preem's profitability because of pure propylene's high market value.

Preem reports it initially tried to use traditional model-predictive control (MPC) for the PP splitter to drive its process to optimum performance and profitability, while still respecting all equipment constraints. However, reliable plant step test data, usually used to create an MPC design, was very difficult to obtain because of excessive settling times and disturbances that prohibited the PP unit from reaching a true steady state. So, Preem's engineers decided to develop the MPC model from data generated by a dynamic simulator instead of the actual process.

Now, it's important to remember that whatever applications that dynamic simulations end up serving these days, their starting point must still be a sound steady-state simulation.

For the PP project, a steady-state Hysys model of the splitter, heat pump and ancillary equipment was reused from a previous study. The Hysys model's predictions of column temperature profile and other variables were validated against plant data. Next, a dynamic simulation was constructed using AspenTech's Hysys Dynamics software by specifying added engineering details, including pressure/flow relationships and equipment dimensions. All basic controllers also were built in the model, which was then checked for consistency and calibrated against process data.

The dynamic simulation was configured to automatically run a sequence of events and record all selected variables. Step tests were then run with Hysys Dynamics, and the data was exported to the DMCplus' model application. Once the step test data was imported into DMCplus, the task of dynamic model identification took on the appearance of any other DMCplus project. Historical process data was also imported into DMCplus to validate the dynamic models. Finally, a DMCplus controller was built and a connection to the DCS was established. Once online, this Hysys-enhanced controller helped reduce propylene variation by 50%, saving Preem more than $55,000 per year.

Because of the extensive simulation effort—and the fact that the underlying model of the PP splitter was sound—no added controller tuning was necessary during commissioning, according to Dr. Nicholas Alsop, Preem's APC manager. The design procedure using dynamic simulation as an alternative to models from plant tests was validated. In addition, the Hysys Dynamics model was next used as the engine for an operator training simulator (OTS), further increasing the value of using dynamic simulation for this study.