Software Architecture for APECS
Optimizing Performance
June / July 2010
Volume 8 Number 3
Sometimes, one of the biggest challenges in transferring a new technology to market is that “the idea is ahead of its time.” That’s how Steve Zitney, director of the Collaboratory for Process and Dynamic Systems Research at the National Energy Technology Laboratory in Morgantown, W. Va., describes the Advanced Process Engineering Co-Simulator (APECS) that he and his colleagues, along with industrial partners Aspen Technology, ANSYS, and ALSTOM Power, have been working on for the past ten years. The key word here is “co-simulator”—APECS combines the powers of process simulation and computational fluid dynamics (CFD) into a single co-simulator that provides the most rigorous, optimized-performance models of complex process/equipment systems possible to date.
APECS helps engineers to optimize the performance of a process or energy plant before construction begins by “building it in software,” thus eliminating many iterations of costly pilot or full-scale plants along the way. Whether it is used to modify the configuration of an existing pulverized coal-fired power plant or to model the next-generation integrated gasification combined cycle plant that will power our future, APECS can provide engineers with the ability to create virtual prototypes more quickly, more efficiently, and at less cost than ever before.
“The unique thing about APECS,” Zitney says, “is that it’s really bringing together two different engineering disciplines and tools. With process simulation, you’re looking at things at the whole plant level, solving material and energy balances for a whole power plant or chemical plant.” Though that sounds complex, it is relatively simple computationally. Because the engineers are only looking at zero-dimensional models, process simulations can be completed in minutes. In contrast, “CFD is a tool that you tend to use for very detailed design of a single piece of equipment, like a gasifier or a gas turbine combustor,” Zitney says. “There you’re not only solving heat and material balances, but all the fluid flow issues as well—flow and mixing issues—so you can see what’s going on inside of a given piece of equipment.” Because of the use of two and three-dimensional models, CFD is very computationally expensive, with simulations that can take many hours or more. Zitney calls CFD “a big hammer” in terms of computational tools.
So how do you combine such disparate engineering tools into one product and sell it to a customer base dominated by engineers who are used to doing process simulation in minutes? (Zitney estimates that a large chemical company may have hundreds of engineers doing process simulation and only a handful doing CFD.)
First, you assemble a team of companies with strengths in both areas. In 2000 DOE began funding the APECS project through NETL. The partners were Aspen Technology and FLUENT. (ANSYS acquired FLUENT in 2006; the CFD product is now called ANSYS FLUENT®.) In addition, ALSTOM Power developed early APECS co-simulation cases and validated results using equipment and plant data.
Next, you make it as easy as possible for potential customers to use your product by supplying plug-and-play code with a standard, open interface so it can be incorporated into commercially available process simulators. The APECS team did this by adopting software standards that were established by a consortium of organizations in the process industries in the 1990s. This made it possible for CFD users in the process and energy industries to easily use their existing CFD models of a particular—perhaps proprietary—piece of equipment in compliant process simulators.
FLUENT soon included a feature that made the interfacing of the customer’s existing CFD models with an APECS co-simulation a quick and systematic procedure.
Though these improvements moved APECS closer to the customer’s expectations, “we realized that in order to really push this technology in the process simulation community we needed to do something about reducing co-simulation time,” Zitney says. CFD is necessarily a slow process compared to process simulation because of the great detail and distributed parameters involved.
And because most process simulations require multiple iteration loops through the whole flowsheet to converge upon an optimal solution to a design challenge, engineers encounter the CFD slowdown more than once in each co-simulation.
To solve this time crunch, the team devised two solutions. First, they made it possible to run the CFD model separately on a high-performance computer in parallel connected to the process simulation running on an engineer’s workstation. Second, they developed a tool to generate fast reduced order models (ROMs) based on CFD results. The ROM tool allows engineers to specify the range of values for all CFD model inputs—pressure, temperature, flow rates—for use in a design of experiments (DoE) calculation. The stand-alone ROM tool of APECS then performs CFD simulations based on the DoE results to create a ROM. This fast ROM can then be plugged in to the process simulation in place of a full blown CFD module. ROMs generally take seconds to minutes to complete, Zitney says, bringing them much closer to the time scales that process engineers prefer.
In practice, Zitney and his colleagues are using APECs to optimize the design of next-generation advanced energy plants, like IGCC systems. ALSTOM is using APECS to model the boiler of a 30 MW pulverized coal plant for municipal electricity generation. They model the plant using Aspen Plus as the process simulator, then model the boiler separately as a 3D FLUENT CFD model with approximately 40,000 computational cells. ANSYS is marketing and selling APECS with their Engineering Knowledge ManagerTM, which manages all the FLUENT CFD models and the ROMs.
Their current strategy is to approach companies that have both CFD experts and process engineers, and show them that they can efficiently combine their existing fluid flow models and process simulations to improve their design process. “We’re saying, you’ve spent considerable time and effort to develop these high-fidelity CFD-based equipment models,” Zitney explains. “We can reuse all this knowledge and leverage it by getting CFD models into the hands of your process engineers to use in optimizing overall plant performance with respect to important fluid flow behavior.”
Tim Palucka is a science writer for NISC, an IBM company, which is a site support contractor for NETL.
Copyright © 2012 | Innovation America