In many ways, modeling has become an established tool for process engineers. Steady-state simulations are now routine for much process design work, while the more-challenging world of dynamic simulation has virtually spawned the rapidly growing market for operator training systems (OTS’) for everything from tightly integrated petrochemical complexes to offshore platforms. However, few would question that dynamic simulation could play a greater role in process engineering. The question is whether it is now poised for wider adoption by practicing process engineers.
The ability to mathematically model a process and its unit operations from first principles arguably dates back to the advent of the first computers powerful enough to do the number crunching — but those mainframe and VAX days are long gone. “Most of us in the past in the chemical industry had our own simulators, before they became commercially available,” recalls John Pendergast, senior technical leader with Dow’s Engineering and Process Sciences Laboratory in Midland, Mich. But the processing and modeling times involved then were hardly suited to dynamic simulations. “Realistically,” he says, “the ability to solve dynamic problems is only about three to four years old.”
Pendergast and his team work with the suite of simulation software packages from Aspen Technology, Cambridge, Mass., on a wide range of troubleshooting and operational issues. “There are lots of unit operations that are inherently unsteady state,” he explains, “that have to be merged into our inherently steady-state simulations — pressure swing adsorption units and batch reactors, for example.”
AspenTech’s senior product manager for both dynamic and batch simulation products, Glenn Dissinger, plots a similar timeline. “The use of dynamic simulation has grown significantly over the last six years or so,” he says, “and it has also become more closely integrated with steady-state tools.” He estimates that around one-fifth of AspenTech’s simulation customer base is using its dynamic solutions, with the popularity growing because of their “ease of use and the ability to tie dynamic simulations back into the steady-state solutions.”
“Ease of use,” of course, can be a somewhat relative term and it’s probably fair to say that dynamic simulation software has not always been noted for its user-friendliness — at least by the general user rather than the simulation specialist. “Even if you have a really good steady-state simulation,” says Pendergast, “porting it over to a dynamic simulation is still by no means at the level of the casual user.” Ben Weinstein, section head in Procter & Gamble’s modeling and simulation department at the company’s corporate engineering laboratory in West Chester, Ohio, makes a similar point. “The primary users of dynamic simulation from my perspective,” he says, “should be the control engineering group, but even good simulation platforms for dynamic simulation are still too difficult for control engineers to use. The big challenge is to make dynamic simulation easier to use by non-simulation people.”
If that’s so, then it’s a challenge that the software vendors are addressing. One of the growth areas picked out by AspenTech’s Dissinger is just that type of integration into the control field of the company’s dynamic simulation products: Aspen Dynamics, Aspen Custom Modeler (ACM) and the Aspen HYSYS Dynamics software it acquired with Calgary, Alta.-based Hyprotech in 2002. (The Hyprotech software was subsequently bought by Honeywell as part of a U.S. Federal Trade Commission (FTC) settlement in 2004 aimed at spurring competition in simulation, although AspenTech retains full development and user rights to HYSYS and the other Hyprotech products.)
“We’ve been working for a few years now on tying in our Aspen DMCplus model-based multivariable control package with our simulation tools,” he explains. In addition, AspenTech also recently released a new linear-state-space matrix controller, which can also link back to the rigorous non-linear models developed in ACM and Aspen Dynamics.
Another addition to the AspenTech dynamic family expected in 2006 is an extension to the Aspen Simulation Workbook introduced last year for steady-state simulations. This basically is a package that allows less-experienced users to translate their process models into the actual simulation model by quickly developing customized Excel-based front ends, rather than immersing themselves in Visual Basic coding.
Meanwhile, Honeywell in the past year has been incorporating HYSYS into its UniSim line of simulation products. The latest release in the range, UniSim Operations Suite R200, came out last December and is effectively an OTS that can also be used for control system checkout. Peter Henderson, London, Ont.-based product manager for the UniSim Operations Suite, acknowledges that the reason for Honeywell’s interest in dynamic simulation was primarily to support the OTS business <em dash>—<em dash> the type of service business that AspenTech, under the terms of the FTC settlement, is excluded from until the end of next year.
“The dynamic models in an OTS have to look reasonable at start up and under normal operating conditions,” Henderson explains. “And the path between those points has to be credible and follow a natural curve, without any mathematical non-linearities or discontinuities. Remember, an operator will see all this in his trend windows and if it’s not what he might expect, then the OTS system loses credibility.”
As well as the HYSYS offerings, the UniSim Operations Suite can also call on dynamic models from Honeywell’s own Shadow Plant and OTISS simulation engines in its Experion PKS process automation system. Shadow Plant, in turn, incorporates dynamic models produced with the gPROMS software of Process Systems Enterprise, London, U.K. Henderson notes that work with PSE started before the HYSYS acquisition and was driven by a desire to take costs out of the development of new unit-operations models. “The formulation and research time and effort is still the same, as with any good model,” he says, “but the time to turn that into an object that can be embedded into a flowsheet is very rapid with gPROMS.”
PSE’s managing director, Costas Pantiledes, characterizes operator training as “a very mature, very steady” market for dynamic simulation. However, he questions whether there is yet a demand for “simple” dynamic simulation. “It obviously has evolved and is being used, but it’s not on every engineer’s desktop and I doubt it will be, as I don’t see the demand for that.” The real value of dynamic simulation, in his view, comes from the complex, very detailed modeling “which tells you things you won’t get just by staring at the problem.”
As an example, he cites the help given by PSE to the BP team investigating the March 2005 explosion at its Texas City refinery. (See CP, Jan., p. 7, for a summary of the final incident investigation report) Working from first principles and with no prior assumptions of liquid and vapor flows in the raffinate splitter believed to be at the root of the accident, PSE’s consultants used gPROMS to develop a high-fidelity model of the tower that showed, against all expectations, that hydrocarbon liquid did in fact overflow the top of the column. “That was a case of dynamic simulation adding value,” Pantiledes says, “in the sense of fully understanding what happened.”
Simulation specialists developed the Texas City model and they too are finding dynamic systems easier to use. “Technologies like gPROMS,” explains Pantiledes, “let you put together models from first principles over a wide range of operations.” And once those models are developed, they increasingly are being used for process optimization. Traditionally, he says, models would have been built for offline design, but working with partners like Honeywell, PSE has been taking the model across the boundary between offline and online applications.
Cal Depew, dynamic simulation product director for the SimSci-Esscor unit of Invensys Process Systems, Lake Forest, Calif., paints a similar picture of dynamic models being put to much wider use. “One of the key trends,” he says, “is that more and more control engineers are now able to use dynamic simulation, because of the improved GUIs [graphical user interfaces], ‘drag and drop’ and interfacing with packages like Excel.” Traditionally, he notes, models would only have been developed for a single purpose <em dash>—<em dash> process design, control checkout or just for an OTS. Now, he notes, a single model can be evolved all the way from a steady-state simulation in, say, SimSci’s Pro/II program, into a dynamic form that can be of benefit in all those applications.
SimSci’s dynamic simulation tool, Dynsim, is powered by the company’s common modeling environment SIM4ME, which provides a consistent look and feel across all its simulation programs. This allows reuse of the same model through multiple stages of the project, Depew says. “Dynsim is a way of bringing the process control department in with the process engineering department, so they can see how the control design and process design interact.”
Other companies expanding their dynamic simulation capabilities include Chemstations, Houston, Texas. The latest release of its ChemCAD Suite boasts a “snapshot” feature for OTS applications. For dynamic simulation, snapshots of the flowsheet taken at any time during the run can then be used afterwards to “rewind” the dynamic time and flowsheet information.
Another firm, Fantoft Process Technologies, Sandvika, Norway, although more involved in upstream OTS applications in oil and gas, has recently been making inroads into the chemical industry with its D-SPICE dynamic simulator. Jan Eckhoff, president of the company’s Houston-based U.S. operation, notes that Bayer recently has taken out several licenses for the software and Shell has just ordered an OTS system for its Norco olefins complex at St. Charles, La. — the fourth such full-scope OTS that Fantoft will have supplied for ethylene plants.
Eckhoff notes that in the past D-SPICE was considered to be the type of simulator that Fantoft itself would engineer, building the models and then delivering the OTS. But the software now has evolved so that third parties like Bayer can do the modeling and simulations themselves.
Plug and play
The goal of allowing third-party model developments to integrate more easily with the big commercial simulators lies behind the work of the CAPE-OPEN Laboratories Network (CO-LaN), headquartered in Rueil-Malmaison, France, and supported by most of the major simulation vendors and operators such as Total, IFP, BASF and BP. (For more on CO-Lan, see http://www.chemicalprocessing.com/articles/2005/510.html.) Gilles Hameury, marketing manager for simulation and optimization company ProSim, Labege, France, says the CO-LaN “component software” approach allows users to assemble only those components needed, and so to reduce costs.
ProSim’s Simulis Thermodynamics software is a case in point, permitting users to generate their own “property packages” to plug into the simulators.
ProSim recently organized a CO training session for specialty chemical producer Rhodia, Paris, France, which CO-LaN’s chief technology officer, Michel Pons, says is a clear sign that industrial users are “increasingly integrating CAPE-OPEN standards into their software considerations.”
Reactions on the vendor side remain mixed, however.
Honeywell’s Henderson sees it “as a great opportunity for the end user, allowing our customers to integrate their own technologies into the simulations.” Depew says SimSci is committed to making Dynsim compliant with the emerging CO standard for dynamic unit operations once CO-LaN members agree upon it. PSE’s Pantiledes notes, “it’s a useful technology that in a few years may be taken for granted.”
For now though, Pantiledes does express some reservations, saying CO doesn’t really go much below the unit-operations and thermodynamics level. “It makes life easier,” he says, “but it doesn’t help in making the model in the first place.” And AspenTech’s Dissinger thinks that other developments, by Microsoft with OLE, have provided alternative ways of integrating applications and models. However, he emphasizes that AspenTech does support the CO interfaces with all its steady-state and dynamic simulation tools.
The simulator companies might be rising to Weinstein’s challenge to make their dynamic products easier for non-specialists, but there’s still a long way to go before ordinary process and control engineers use such software routinely.