Moving through the maze of engines at Pratt & Whitney's historical museum in West Palm Beach, Fla., manager Mark auBuchon stops short and points out a 13-ton granite block.
"In the past, we set our engine tools here," he says, "so they're on solid, still ground." On these balancing platforms many design predictions were made for jet engines-very complex machines for converting hot, high-pressure air into thrust.
A different kind of balancing act today takes place as industry and government partner to create the first full-system analysis tool. Within NASA's High-Performance Computing and Communications (HPCC) program, NASA Lewis Research Center leads this tool development for the design of aircraft engines called the Numerical Propulsion System Simulation (NPSS).
The NPSS team also consists of representatives from General Electric Aircraft Engines, Pratt & Whitney, The Boeing Company, Allied Signal Engines, Allison, Arnold Engineering Development Center, Williams International, Teledyne Ryan Aeronautical and Wright Patterson Air Force Base.
How to improve?
What is the best process for building engines? That's how team members frame their work as they forge ahead with a tool that could transform America's aeronautic design practices. Currently, building these delicate, complex machines is costly and requires a long time-to-market. "The average development cost for an engine is $1 billion," says NASA's John Lytle. "The time-to-market is three to five years for a commercial engine and 10 years for a military engine."
"People can work their entire career on one engine," adds auBuchon.
It's not always easy to make a change to an established process. Transforming this development process will come as a result of individuals and teams adjusting internal design processes in such a way that savings, speed and quality breakthroughs become not just possible but inevitable.
For that reason, NASA Lewis Research Center rallied the industry to build a common tool for modeling an entire engine. "This model projects engine temperatures, pressures, thrust and noise under various operating conditions," explains GE's Ron Plybon.
Each engine company currently has its own proprietary engine cycle model based on its own standards and data models. The traditional cycle model is a non-dimensional (0-D) representation of the engine, where each component is a box whose performance is represented by a data table or map instead of a drawing. The goal of NPSS is to restructure the existing cycle model to enable analyses from 0-D to 3-D to be performed in a seamless manner throughout the design and development process.
"We plan to switch to NPSS to model all new engines over the next two years," states Plybon, who cited nearly the same timeline as his industry competitors. "When we implement NPSS, we'll see a 50 percent improvement in the way we do business with our partners and our customers (airframe and airline)," adds Joe Osani, who is responsible for managing the business side of GE's NPSS development.
"We're making sure NPSS will meet all our customers' needs," adds P&W's Thomas Mirowski. Nearly three-quarters of his engineering group is devoted to crossing the T's and dotting the I's of this tool.
NASA's role as gatekeeper of the tool-building team is challenging because each company has a different approach and different codes to contribute, says Boeing's Russ Ashleman. Clearly all parties, and especially Boeing, stand to benefit from having a common cycle system. This would mean the same computer program is delivered by engine suppliers-P&W and GE-to the airframe company.
Improving the tools used for building engines will require a new approach from the methods used over the last 30 years. The classic model delivery scenario is for a supplier company to create a cycle deck, which is then delivered to a customer company. The goal is to be able to share models with others while imposing various restrictions about what the recipient can see or change.
They are called decks because in the early 1970s decks of punch cards containing computer programs were exchanged between the airframer and the engine companies. While the computer program has evolved a bit since then, the FORTRAN code has remained the basis for today's modeling. The deck or cycle model is used throughout the engine's lifecycle, from initial investigation through design testing and later as a tool for probing in-service problems.
These computer simulations of engines become the engine supplier's brochure for the airframe company. "The model itself executes very quickly, but it can take months to get the model set up," says Plybon.
"Unfortunately, everybody that sends us a customer model gives us something different," adds Boeing's Ashleman. "It can take suppliers between one and six weeks or longer to give us the models. And then another week before we get it installed, assuming our engineers are able to run it."
Ashleman offered one reason why these decks take so long to process: double compilation occurs during the set up. The suppliers compile at home and at the customer's location.
The long process impacts the make-up of their work. "A large percentage of what we do is not creative," asserts auBuchon. It's the mechanics of moving information." "NPSS," he says, "moves the drudgery to the computing systems, freeing engineers to do the fun things."
Subtract the setup time for readying traditional decks, and design iterations become easier. Design changes between the airframer and suppliers can be challenging without fast turnaround. This is true especially for first-time engines. "Years ago, the 777 program required new, higher-thrust engines for brand new airplanes," states Ashleman. "So in that case, we did several iterations. The engine company did its own development program and sent us data. Then we again asked for more thrust. Meanwhile, your airline customer says, 'Well, we want to fly this plane between these points and carry this number of people.' So we iterate those design criteria into the process too."
NPSS is the key to moving models quickly among suppliers, airframer and customers for these iterations. "The vision for NPSS is to be a numerical test cell that enables full-engine simulation overnight on cost-effective computing platforms," says NASA's Cynthia Naiman, who manages a five-year timeline to reach that vision.
Not only is the speed of making changes necessary for design, but also for testing. "We have a requirement to rapidly adjust our engine model so it matches the test data and still maintains the overall characteristics and accuracy of the engine model," says Plybon.
P&W's Mirowski, who has performed benchmark tests comparing Pratt & Whitney's in-house simulation to the NPSS simulation, says that NPSS allows unlimited testing of the "whole flight envelope, which is not possible when testing the real engine. That builds our confidence in the engine design." This unlimited analysis is a result of NPSS's enabling a wider range of high-fidelity analyses than the traditional 0-D cycle model.
With this ability to zoom to different fidelities, "designers can try potential changes on the computer without building and testing the hardware," adds Mirowski. "If you have to go through an optimization process that builds 10 different pieces of hardware, that's tremendously expensive. That's a trial and error method. NPSS offers a big savings boost by providing more detail early in the design process."
"Most engine and design costs are locked early in the process when you're using cycle decks," adds Plybon. "You only achieve detailed models of the engine after you've already committed most of the cost. At that time, it's difficult to go back and make changes."
NPSS tools will allow higher-fidelity tools to be used when there is opportunity to change and impact engine design. "Currently we execute a 0-D cycle model, generate some information and pass it along to a specialist who runs the higher-fidelity code," states GE's Plybon. "NPSS will allow us to perform detailed system-level analyses that are usually carried out only at the component level."
On the other hand, "it provides a more efficient way of doing some of our current low-fidelity analysis while allowing access to higher-fidelity codes," states Plybon. "Now, with NPSS, instead of taking months to move high-fidelity codes in a model, it could take days because this tool is set up to receive them."
"What you want is to quickly combine the lower-fidelity information with the higher fidelities in specific portions of the design. This is what we need to do to take this modeling system to the next step," agrees P&W's John Mason.
Access to higher-fidelity analysis is just one of the value-added capabilities of NPSS. A modern computer architecture and the distributed, object-oriented capability are also new features in the NPSS tool.
Industry teammates cheer the move to object-oriented Common Object Request Broker Architecture (CORBA), a software package built into the tool. With this interface the NPSS team can easily collaborate by taking the legacy codes and cloaking them in a CORBA wrapper or calling routine. "The routine will pass on the appropriate information and tell it to run; it runs and then passes along pressures, temperatures and other information to the rest of the model," explains Mirowski, who is leading P&W's transition to NPSS. "And because it's an object-oriented structure, it's easier to manipulate while appearing the same to the NPSS tool."
Also planned for the future in NPSS is the integration of multidisciplinary design. An early pioneer of NPSS, NASA's Greg Follen attributes NPSS's current momentum in the industry, not so much to its technology but more to a successful strategy for combining disciplines such as aerodynamics, structures and heat transfer with numerical zooming on component codes. He says, "As these combining disciplines add input to the overall simulation, the effects of these disciplines can be measured against the geometry, the loads, the weight-all at once-as opposed to a serial process of passing along the analysis to each developer."
Boeing hopes to interface NPSS with current in-house codes that are used to measure airplane performance and run them concurrently, adds Ashleman. "Our structure, noise and emission staffs use output from these cycles. Combining their programs with NPSS would add capability and faster turnaround."
First things first, says auBuchon. "Let's get the plug and play of the whole toolbox full of 0-D models. Then within five years, we will release a true, high-fidelity NPSS simulation."
"The government's role is to act as a catalyst for commercial products," states Lytle, who is among the early NPSS visionaries. "Once the final software is out in the community, NASA backs away. Then the industry consortium will maintain it."
And NPSS may serve an estimated 600 engineers and scientists in the engine community. Ashleman joins the rest of this consortium in recommending the tool for other applications: "We're talking about using this tool system to model rockets, too."
For now, entering a new paradigm of design and engine validation is a bit challenging. An early release of NPSS was distributed only last year. As the next release nears its tollgate in September, engine companies are confronting the fact that this new tool could restructure their organizations.
"That's not so difficult for P&W; we are a change mechanism. Many in this industry are organized by having the employees grouped by their respective engine component. This is because historically they don't have the capability to analyze the entire system with high fidelity," auBuchon says.
What is somewhat difficult is the NPSS teams' patience with everyone involved in building a common industry tool. "What we know is that we cannot do it by ourselves; it's bigger than we want to attack alone. But when we join together in unifying our multi-disciplined design practices-leveraging our investment in our own proprietary systems-we achieve a different level of capability. We can take that tool into the world and utilize it for all that we do in design and development," says auBuchon.
GE's Plybon adds that the development of this tool is an excellent marriage between NASA, with its futuristic goals, and industry, with its economic goals.
And this new approach promises to benefit everyone with safer, cheaper, more reliable systems. It also may address man-made global warming. "We don't have a lot of tools for that right now," says Ashleman. "I think NPSS will provide a framework for developing a good emissions model."
While clearly cooperating to form a common foundation, industry will compete by the type of analysis individual entities choose to do with NPSS, says auBuchon. "Whoever exploits this new technology to the highest degree wins."