Ian Sheppard/LONDON
Massively parallel computing, or "hypercomputing", methods which are expected to revolutionise the design of future aircraft and missiles, including the next-generation high-speed civil transport (HSCT) and the US Tomahawk land-attack-missile replacement, are now being developed.
NASA's Conceptual Design of Air Vehicles Hypercomputing and Design project was set up as a technology-transfer initiative to exploit advances in computational science, artificial intelligence (AI) and modelling/simulation. Industrial participants include Aerospatiale, Boeing, General Electric, IBM and Lockheed Martin.
As part of the initiative, the "Vizlab" at the State University in Rutgers, New Jersey, is undertaking research into the conceptual design of inlets and exhaust nozzles, focusing on AI-augmented performance optimisation.
The Rutgers team uses "modelling constraints" to fine-tune designs. The team has managed to design automatically HSCT airframes and exhaust nozzles, using a "multi-level" AI approach, where an abstract "design space" is used to reduce the design possibilities at a more defined low-level design space, thereby saving time.
The researchers, led by Professor Doyle Knight, claim that it is the first successful combination of numerical optimisation methods with Reynolds-averaged Navier-Stokes (RANS) fluid-dynamics simulation for high-speed inlets and nozzles, and will lead to an "order of magnitude" reduction in design costs.
Rutgers is collaborating on the project with the United Technology Research Center, which has applied the AI-augmentation optimisation technique to the design of an inlet for an inexpensive supersonic missile to replace the US Navy's Tomahawk, and dubbed "Project Cheapshot".
The team developed a supersonic mixed-compression missile inlet with fixed geometry and no bleed air, with a cruise speed of Mach 4. It achieved a pressure recovery of 0.409, 32% better efficiency than when using two-dimensional design methods (developed for the comparison with that of the industry-standard, standalone RANS techniques. Some NASA hypersonic-inlet designs were also redesigned, and performance was found to have significantly increased in the simulations.
Rutgers claims that industry does not use automated optimisation in inlet design, but has shown that the benefits are considerable.
Another application is for when slower boundary-layer air is "bled" from the inlet to prevent the flow breaking down and causing the inlet to "unstart". Traditionally, this design has been too complex to model successfully, but the Rutgers team is now attempting to design an inlet which operates at high angle-of-attack and is far more tolerant of gusts.
Source: Flight International
