(Based on information supplied by Bill Feiereisen, NASA)
The Aerospace Engineering field is in a sense a revised Grand Challenge with a business model having the following characteristic stages:
There is a tradeoff between how good or efficient the system is (in terms of $'s or performance) and the time to market. Earlier delivery generally captures a greater market share and is the main key to greater revenue. Shortening the design process by six months will typically capture much more of the market - potentially $0.5B for a typical airliner. Delaying the lines freeze, eliminating some steps (such as the pressure model) also help and this is where simulation, and HPC can make a difference. There are however problems with proprietary codes and problems with proprietary data outside of the companies direct control and many security concerns. This means a lot of the work must be carried out internally to the corporation and makes it difficult to transfer new research technology into the companies in a timely fashion.
Currently, the design cycle in bringing a large aerospace systemn to market involved running large CFD codes with a some structural engineering input. Typically this would draw upon the disciplines of four grand challenges oriented around four types of vehicle. Relying solely on computation for all its development is not a trusted option. This issue of trust is entirely separate from computing capability and has to do with the numerical algorithms and significantly turbulence modeling. The current design process is not one big CFD code but rather a combination of many CFD calculations and experiments. Typically no simulations are run concerning manufacturing or maintainance which is unfortunate as these factors are the really big drivers in cost, as well as the cost of capital itself.
Improved operational cycles will combine experiment with computation and will involve CFD computation at all levels. Collaborative technologies will contribute substantially in the form of: remote access to wind tunnels; remote control of experiments; and remote access to real-time data. Access to large experimental (and CFD) data sets for analysis are expected to lead to cooperative relationships. This industry is very risk adverse, and companies cannot afford to depend upon a process that might not work. Government sponsored work may be helpful in developing processes that are more efficient but which have a greater risk of failure.
Most of the aerospace companies say that they will not buy the next generation supercomputer, since it is perceived as too risky an investment. Instead, companies are exploring networked workstations and some such as Pratt & Whitney and MacDonnell Douglas, have shown great success.