Absolut lifecycle
A new distillery for vodka benefited from simulation at every stage, from design to operator training

Process modeling gives the greatest value if a single model can support steady-state and dynamic simulation, plus specialized applications such as operator training, and has the ability to act as a data repository. Hysys from AspenTech helped Swedish vodka producer V&S Absolut Spirits carry out all these functions for its new distillation plant.

Absolut Vodka is a premium brand, so the new distillery being built by V&S Absolut Spirits at Nöbbelöv in Sweden has to incorporate the best in process design and control. Simulation based on
Hysys software from AspenTech has allowed engineers to get optimum results from the new plant at every stage from concept to operation. Steady-state Hysys models were used to help design the distillation process to ensure that vodka from the new plant would meet the legendary purity standard of the Absolut brand. The same models were also used as the basis for dynamic simulation used to verify the controllability and operability of the plant.
Once the optimum design had been chosen, the team also used the dynamic model to create a virtual plant simulator (VPS). Connected to distributed control system (DCS) consoles identical to those used on the real plant, the VPS provides a realistic environment for training operators, and helps engineers develop and verify plant operating procedures. At every stage, the Hysys model has acted as a central database of engineering information that is continually updated.
Although vodka is almost pure ethanol, it contains many trace components whose concentrations must be accurately controlled. The company chose to work with Hysys primarily because of its ability to model the six-column distillation process accurately. With a variety of add-ons, the software can also be used at every stage of the project’s life cycle.
Designing a better process
The first step in designing the new distillery was to investigate the existing one. The designers needed to know how the many process variables affect product quality. To minimize the risk of creating off-spec product during these tests, they used the software to create a model of the process.
The model’s ability to predict changes in variables that could not easily be measured was essential in building up understanding of the distillation process. After six months of fine-tuning the model, the team could accurately predict the effect of changes in every process variable, and
were confident that they knew enough to design a new plant.
In drawing up the flowsheet for the new distillery, the aim was to balance design simplicity against capital costs, operating costs and controllability—without compromising product quality. Steady-state modeling was a valuable tool at this stage.
Extensive recycles and heat-exchange networks can cause controllability problems, or make startup and shutdown difficult. To check the dynamic behavior of the plant, the team used their steady-state model as a starting point to build a dynamic model. To ensure consistency, the dynamic model uses the same physical properties package and the same 16 components as the steady-state model. Models were validated against a steady-state heat and material balance to an accuracy of 65 percent of instrument span for the key variables, or 62 percent for critical process streams.
The main objective was to make starting up the plant as quick and trouble-free as possible. Engineers used the dynamic model as a tool during risk evaluation sessions, imagining potential problems and then checking their effects using the simulator. Some important lessons emerged. For instance, the original control strategy for the raw spirits column was to
have the level in the reflux drum control the reflux valve, while the product takeoff valve was to be controlled by the composition analyzer. However, the model showed that this strategy made it difficult to control the overhead purity. As a result, the control strategy was changed to use the reflux valve for composition control and the product valve for level control. This change introduced another problem. During startup, the operator now had to manually control the reflux drum level, and the presence of a high level switch in the reflux drum raised the possibility of tripping the plant. The team devised a way of using the reflux valve for level control during startup, and then smoothly changing over to the normal operating configuration
once the overhead product reaches the correct purity.
The steam-heated reboilers were modeled to study the conditions under which
live steam breaks through to the condensate system. This can happen if the load on the reboiler drops suddenly, or if the temperature controllers demand more steam than the reboiler can condense. The team was able to develop a strategy which detects when steam is about to break through and tries to prevent this by
backing off on the flow controller signal.

A better training experience
The behavior of the large centrifugal steam compressor used for energy recovery was studied, especially during startup. The compressor supplier and the Absolut team were able to work together, using the dynamic model as a basis, to develop a
safe and effective startup procedure.
The model also acted as a repository for plant data: dimensions of pipes, valves, pumps, columns and tanks, flowrates, temperatures, pressures and concentrations. Visual Basic macros were used to export this information to Excel datasheets that could be used to specify equipment from vendors.
Even on a modern, automated plant,
making vodka of high and consistent quality depends on the skill of the operators. Skill implies experience, and experience takes a long time to acquire on a real plant.
To train operators better and faster, the team created a virtual plant simulator (VPS) based on the Hysys dynamic model. The VPS is a realistic, risk-free environment in which operators can practice running the plant under normal operating conditions, transients such as startup and shutdown, simulated fault conditions and emergencies. It also helped the engineering staff improve their understanding of the process, design the operator interface and fine-tune the plant’s operating procedures — all before the real plant has been started up. Six instructors use the VPS to train the plant’s 15 operators. Each instructor uses a dedicated workstation to control the training session, select the plant operating mode, introduce “failures” as required, monitor the trainee’s performance and produce a detailed report based on data logged during the session. The table on page 38 shows typical malfunctions introduced by the instructor.
The “backbone” of the VPS is AspenTech’s Music simulation framework. This links the process model, created in Hysys Dynamics, with a simulation of the DCS and workstations for the instructor and trainees. The ABB DCS is simulated using ABB’s AC800F emulation system. This uses the same database configuration files as the real DCS. Data is exchanged with the Hysys-Music simulator via OPC standards, and a custom API supports functions including Save/Restore Initial Conditions, Save/Restore Snapshots, Freeze/Resume controllers and time synchronization. All the VPS hardware, including the ABB OperateIT consoles and network, is based on standard PCs running Microsoft Windows 2000. This reduces the hardware cost of the VPS by a factor of ten or more compared to simulators based on other DCSs that require proprietary hardware for their operator consoles. The operator stations use ABB’s AC800F OperateIT system to provide a graphical interface that is identical to the one on the real plant.
Preparing for startup
First reactions from the operators were that the VPS acts almost exactly like the real plant. Once the engineering team had set out the startup sequence, the operators were able to begin practicing. There was even something of a competition to see who could start the plant and get it stable in the shortest time.
Subsequently the VPS will be used to train new operators and maintain the skills of experienced operators. The plan is to have more consistent product quality, fewer unplanned shutdowns and a team of operators who can handle most incidents without having to call for backup.
Apart from operator training, the VPS has many other uses. For instance, it has helped to ensure that the configuration of the DCS is free from logic errors and bugs. It also provided a way to check the alarm settings and alarm loading of the DCS, so reducing the number of spurious or redundant alarms that the operator has to face when something goes wrong. The VPS was also used to derive a set of preliminary tuning constants for the controllers. Absolut’s control engineers spent a lot of time optimizing the tuning constants using the simulator, with the aim of cutting the time needed for loop tuning during commissioning.
For the future, the VPS will be used as a knowledge base, storing information on all plant incidents as they occur, and the corrective actions taken. Operators who were not on the scene at the time of an incident can replay it later and try their hands at controlling the disturbance. The VPS will also be used to check future changes to the DCS database before they go live, and as a basis for designing advanced control systems.
The involvement of the Absolut team was a key factor in the success of this project. Kirk Pitts, the Aspentech modeler, commented: “I am amazed at how deeply involved they were in this project. I’m
sure that because of their involvement, this was much more valuable to them than just being a training platform for their operators.