Through the use of computers and highly sophisticated computational methods, Lockheed is taking major steps toward meeting the ultimate challenge of stealth -to design and field an aircraft that remains undetected at any range and at any altitude. In the not-so-distant past (1960s and most of the 1970s), stealth design was primarily based on experience and guesswork. Since that time, we have begun to rely more and more on computer models. These models, combined with supercomputers, have allowed us to better understand the cause-and-effect relationships of stealth and, hence, to produce better designs. In the context of stealth as an art, supercomputers have clearly increased the size of our canvas, while the highly developed software tools have allowed more imagination and creativity to enter into the design.

Lockheed SR 71 Blackbird

The practical benefits of having the computational approach play a more significant role in stealth design are numerous. For example, we can examine many more design options in a given amount of time, and we can also reduce many of the risks in going from novel concepts to successful hardware. This approach, in the concurrent engineering sense, can lead to reduced costs, while insuring that the "-Illties" (manufacturability, reliability, maintainability, and survivability) are satisfied. But, we cannot simply let computer power go to our heads. The reality is that stealth is not cheap, and in order to make the best trades with respect to cost, "-ilities," and stealth, we believe the computotional environment is the best place to do this.

The ultimate challenge is, therefore, to meet all of these requirements, and to do this we will most certainly need our best resources, including new powerful computational tools. We will also need to use them in a concurrent engineering design environment. ln this environment, the fidelity and, especially, the speed of the computational modeling process will need to be as great as possi 'ble. Only then can we search the entire design space and have any chance of finding the "absolute best" design.

Particularly critical to stealthaircraft design are the external shape, structural composition and electrical properties of materials used, sensor apertures, surface breaks, engine inlets and exhausts, and weapons carriages-to mention only a few. The diversity of possible design combinations is countlless, and experimental approaches to achieving an optimum design are likely to come up short. On the other hand, with supercomputers and highfidelity computer models, these experiments can be done numerically, and many more ideas and concepts can be tried than would be possible to build and test in a lifetime.

Now that the case has been made for the role of computational methods in stealth design, we will next present a brief history and current status of these methods at Lockheed. Finally, we will describe the importance of computer graphics, automatic steakth design, and the utility of visualization in the design process.

Supercomputer boom of the 80s and 90s - Cray, Connection Machine, Intel

Aircraft design is an Iterative process requiring inputs from many disciplines. In the past, this process was largely based on experimental methods whereby expensive hardware models were built, tested, modified, and retested many times. The end result was often determined when the money ran out, and the design was likely to be sub-optimal, based on the available technologies at the time. This was a high price to pay, but alternative methods were not available.

CAD model graphics workstation - Supercomputer.
(Solid model, top and mesh model)

Aircraft are very complex structures, and their radar signatures are not necessarily related to these structures in a simple way. Until the late i98os, the computational models available could not adequately model the hardware being built. As an example, no algorithm was available to predict the radar cross section (RCS) of a complete, three-dimensional, detailed vehicle with radar-penetrable materials.

Our ability to predict the RCS of vehicles has actually tracked closely with the evolution of the computer. In fact, the argument can be made that the limitations of computing hardware have been the shortfall for computational methods all along. Toward the end of the I970s, scientific computational capabilities began to improve (e.g., IBM 370); however, the available computer models were only limited high-frequency approximations and turnaround times were still unimpressive. For example, very simplified, facet models, (e.g., no control-surface break effects or engine-inlet effects) took us up to a week to compute and assemble for useful results. No interactive design was possible even though simple analysis algorithms were used. Often the RCS of inlets, exhaust nozzles, and sensor apertures were measured experimentally and added in piecemeal. No accurate numerical models existed for computing RCS from these realistic features, let alone from the complete model. The result: the design process was time consuming and expensive.

The most sophisticated computational models available inside Lockheed in the late 1970s were threedimensional, but they were limited to all-metallic bodies and faceted approximations to curved surfaces. These computer models predicted entire vehicle signatures, but they were only approximations to the wave phenomena, and they were accurate only over a limited range of radar frequencies. The basic approach was to break the vehicle into its component parts (including facets), compute the radar signature of each part, then add these individual contributions to achieve the final result. Other so-called exact theories were known, but the computer hardware aval 'lable at the time was not even close to being large enough or fast enough to solve problems of interest to Lockheed.

Lockheed's present computational capabilities grew significantly during the i98os. It was not accidental that this development was coincident with the arrival of supercomputers for largescale scientific computing. During this period, Lockheed developed its highfidelity, 3-D, radar-signatureprediction tools for complex shapes with nonmetallic and radar-penetrable materials. Hence, these models predict the RCS of a complete stealth aircraft. They have been optlmlzed for the multi-processor vector Cray supercomputer and have been designed and verified to be on par with experimental methods.

Visualization of computed results

Computer graphics is another technology that grew out of the i98os and has now become key to stealth design. Computer graphics provides the means to describe geometric shapes in a context that both the computer and the RCs engineer can understand. It also provides a platform for display of the computed results. The graphic images seen in the epic movie Star Wars attest to the sophistication of these capabilities. This same detail (and more) is needed to describe a stealth aircraft and achieve the accuracy and fidelity needed for signature calculations.

Visualization of computed results

To see the utility of computer graphics in RCs design, let's follow the process. The geometric data that describe the physical features of the vehicle are obtained from a computer 'ded design (CAD) database of vehicle characteristics. This same CAD database is used by the aerodynamicists, the propulsion engineers, and others, and eventually the same database becomes the basis for generating drawings for manufacture of the vehicle. Special graphics software and graphics workstations are used to view and modify this CAD data and to prepare it for the computational task ahead. The complete computational model that goes to the Cray is the geometric description of the vehicle and the set of mathematical equations describing the behavior of electromagnetic waves as they interact with the vehicle. The model (code) is then executed and the results are returned for display on the same graphics workstation.

Enhancements in RCS prediction models can be expected in the future, but the most significant improvements in stealth designs will come with the arrival of next-generation computing hardware. Limitations inherent in the present vector-multl-processor approach will persist, and only modest improvements in this style of computer can be expected over the next two to five years (e.g., estimate two- to threetimes increase in size and speed). As an example of this limitation, a calculation Of RCS for a simple delta-wing aircraft with a 50-meter wing span requires, at the very least, nearly 96 billion memory locations (one alphabetical character equals one memory location) of (disk) memory and approximately io days to compute on the largest, fastest production Cray now available (YM-P/8128). Also, if design changes are made, another io days is required before the results would be available.

This problem may seem large, but it is typical of the size problems we need to address. It involves more than io,5 floating point operations (calculations). The practical impact of such intensive calculations is both good and bad. The good news is the recipient of the data gets his money's worth, but the bad news is the computer users who are displaced have to wait to days to reaccess the computer.

However, a solution to this computepower problem appears to be on the way. The U. S. computer industry, major U. S. universities, and the Defense Advanced Research Projects Agency have been working together since the mid-1980s to develop parallel supercomputer technology, and the results of this collaboration have begun to show the enormous gains that were expected at the outset. In fact, the newest generation parallel machines are capable of one-day turnaround, at most, for the above problem.

At present, no automatic RCS design capability by computer alone exists at Lockheed. However, such techniques are currently being developed and we expect to begin using them by 1995. Since the essence of successful stealth design resides in the heads of the RCS engneers, this automatic design package must be knowledge-based, i.e., the design rules that guide the design process must be derived from the design engineers themselves. And, since technology is never static, especially lowobservable technology, these rules will continue to evolve. The RCs design engineers will remain in-the-loop to fine-tune the rules with small changes, and to come up with completely new rules that reflect major breakthroughs in stealth technology. Computational models, like those mentioned earlier, and next-generation parallel supercomputers will be key to this automated design process. After each "suggested" design change, the RCS calculation is repeated to verify that signature reduction (or no signature reduction) was achieved, and the process is repeated 'l an optimum design is found. The untl I 1 need for raw speed in the computer is clearly the "long pole in the tent." The design changes and the optimization itself can be made subject to any given set of constraints, including cost and changes that impact the "-ilities" of the vehicle.

The major question now is, "How will the RCs engineer derive the new rules for stealth design?" Lockheed believes that part of the answer lies in what data the engineer chooses to look at, and how it's displayed. This data may be a computational result from a numerical experiment, or it may be from a range experiment where a model was actually built and tested. ln either case, a large array of data is available -in fact, often, too much data to comprehend.

This is where computer graphics can shine. For calculated or experimental results that are steady (periodic) with respect to time, these results can be integrated (postprocessed) with the original geometric description of the vehicle to reveal steady-state wave phenomena that control radar signature. This imagery can be as important to the RCS engineer as an x ray is to an orthopedic surgeon.

Advances in computational electromagnetics - Visualizations of wave phenomena computed on a parallel supercomputer

If the wave phenomena being simulated by the computational model are for the transient case, instead of the steady-state case, and the computergraphics hardware is connected to the supercomputer during the calculation, visualization of the temporal results can occur as the computation proceeds. The RCs engineer can observe the simulated phenomena in slow motion and easily study the temporal behavior of the radar "pulse" as it strikes the vehicle. A video-tape can also be made and the animated phenomena studied later. Visualization of this type is still new to the low-observable design community, so little design experience has been accumulated. However, in the future, we expect a new class of low-observable design procedures (rules) will evolve based on this moving picture of Maxwell's equations (the equations of physics that describe electromagnetic waves).

VAUGHN P. CABLE was appointed Senior Research & Development Engineer for Lockheed Advanced Development Company in December 1990. In this capacity, Cable works with the technology division in examining the observable characteristics of a variety ofplatforms. His duties include supercomputer modeling to assist in the application of electromagnetics to the design of low-observable vehicles.

Before joining Lockheed in 1988 as Chief Scientist - Wave Phenomena, Cable compiled ten years of aerospace industry experience, holding tbe positions of staff engineer at Hughes Missile Systems, senior scientist at Pacesetter Systems, and, most recently, manager of advancedprograms at California Microwave. Prior to that, Cable bad been an assistant professor at California State University, Northridge.

Cable was born in Santa Monica, California, on I3 July, 1943. He earned both a bachelor's and master's degree in engineering from Cal State Northridge before receiving his Ph.D from Ohio State University in 1975. Active in the Institute of Electric and Electronics Engineers and the Sigma Xi Research Society, Cable holds two patents in medical electronics.

© copyrights 1992 by Lockheed Corporation, 4500 Park Granada Boulevard, Calabasas, CA 91399-0610
published by plweb publications - Gregor Mima, Technical University Vienna.