For instance, GenCorp's Aerojet (www.aerojet.com) division in Sacramento, Calif., is using the MathWorks' (www.mathworks.com) MATLAB and Simulink software to build a control system to deliver constant pressure in the fuel tanks of Kistler Aerospace's K-1 space launch vehicle. When finished, K-1 will be the first commercial, reusable launch vehicle with low-cost access to space for low-earth-orbiting satellites. K-1's modified Russian rocket engine from Aerojet burns liquid oxygen (LOX) and kerosene, but as LOX levels in the tank drop, more pressure is needed to force it to the combustion chamber. To restore pressure, LOX consumed during firing is replaced with helium from external storage tanks. However, helium creates a new control engineering problem, because LOX requires steady pressure, while the helium requires constantly increasing pressure.
As a result, Perry Stout, K-1's controls engineer, developed a solution in which flow between the high-pressure helium storage tanks and the low-pressure LOX tanks is controlled by a series of flow-regulating solenoid valves and an orifice that can vary in size with ambient conditions, and used MATLAB and Simulink to design a control system that regulates valve operations and orifice size. First, he derived and wrote out the physical equations, moved these core equations into Simulink, developed and tested a model and graphically added heat-transfer equations and closed-loop control laws without writing added code. Next, he used MATLAB to analyze the design and modified it in search of optimal conditions, a task that would have taken several months using a conventional engineering process.
"If this was a Fortran project, the control system design would have involved an entire team," says Stout. "A manager would have been required to divide the modeling, simulation and control tasks among several people, closely monitor and coordinate all activity and summarize the results."
Jason Ghidella, Simulink process marketing manager for MathWorks, adds that, "Simulators are becoming more dynamic and non-linear because high-end users want better performance and safety. They want to see if their control strategy is in the ballpark, but they also want to throw all kinds of unusual conditions at it too," says Ghidella "Non-linear simulations are based on partial differential equations (PDEs) and differential algebraic equations (DAEs), and these require a lot of calculating that can't be done in traditional ways. So simulator developers must find other ways to understand and solve these dynamic problems and then generate a response."
Start-Up and Incidents Handling
Similarly, simulations are even being used to assist routine operations, as well as evaluate and respond to alerts and events. For instance, Rio Tinto Alcan's (www.riotintoalcan.com) Gove bauxite mine and alumina refinery is located at Nhulunbuy in Australia's Northern Territory, and it just completed a $3-billion expansion that will increase its alumina production from 2 million to 3.8 million tonnes per year and allow the refinery to run independent of its local bauxite reserves. As part of its expansion and to improve plant capacity, RTA Gove chose a new double-digestion technology, which uses low-temperature digestion for removal of trihydrate alumina, followed by high-temperature digestion for monohydrate alumina. Few double-digestion circuits are established yet, so RTA Gove needed a simulator to train its operators without adversely affecting plant operations, and to help its multi-DCS controls architecture avoid data mismatches.
Consequently, RTA Gove picked Honeywell Process Solutions' (www.honeywell.com) UniSim Operations simulator to do months of operator training prior to start-up on its control system that was developed six months earlier. UniSim Operations is a direct-connect, full-replica, dynamic process simulator, which allows a high-fidelity model of the process to run in real time and appear from the DCS console as though a real plant is being controlled. UniSim's software contains a library of modules that mathematically represent the behavior of process equipment, logic and control components under dynamic conditions. The modules include heat and material balances, operating equations, thermodynamics and physical property calculations. These modules are used as building blocks to create a realistic representation of a specific process, area or plant.
Using these tools, RTA Gove's double-digestion process model includes 135 tank modules, 85 pumps and 1037 control valves. There are 386 field-operated devices and 7370 control points simulated. Training features also include 1242 malfunctions. The process model takes about 0.2 cpu seconds to run, and the model runs every 2 sec, which is more than enough to realistically simulate the plant's process dynamics.
"One of the biggest benefits we've received from UniSim is improved operator effectiveness. Like most operating alumina refineries, our equipment is operated continuously, and many operators are not well-practiced in running under start-up, shutdown or emergency conditions," says Manoj Pandya, Rio Tinto Alcan's alumina projects manager. "Similarly, in new installations, operators may have even fewer skills in managing the process and the knowledge of the equipment limits, even under normal operating conditions. UniSim enabled us to train our operators in advance, so they could practice new skills without hindering the plant."