Current Research Projects
Design of Complex Multidisciplinary Systems
A key challenge in the design of complex systems is the interaction among various engineering disciplines. Considerable performance gains are possible when systems are designed in an integrated fashion--when components are shared by multiple systems or when combining systems produces favorable behaviors. However, classical design strategies separate complex systems into fairly arbitrary disciplinary divisions to reduce complexity and to enable the systems design processes to occur in parallel. This research focuses on techniques to design systems by exploiting beneficial system interactions, while enabling parallel design and minimizing the amount of data that must be exchanged between disciplines.
Gradient-based Multifidelity Optimization
More and more, engineering analysis codes provide sensitivity information as part of their solution. Further, many engineering codes now use adjoints to find solutions more quickly and/or more accurately. This research focuses on how adjoint and sensitivity information can be used to speed the process of optimizing complex systems.
Gradient-Free Multifidelity Optimization
Many engineering systems must be designed to meet specifications placed on high-fidelity analyses, for example, computationally expensive simulations or costly experimental validation cases. Many of these analyses do not provide design sensitivity information, and in many cases it is impossible to estimate sensitivity information accurately. When this occurs, multifidelity methods based on gradient-free calibration techniques may considerably speed the design process. This research focuses specifically on methods that can be proven to converge to a high-fidelity optimal design if given a sufficient number of design iterations.
Sample methods are available for download and testing. DISCLAIMER: Technical support is not guaranteed and may be sporadic at best. In addition, there are some known robustness issues, such as using a 2-norm and not an infinity norm for the trust-region problem, and that correlation functions in use are isotropic and not anisotropic.
Past Research Projects
Low-Speed Aerodynamic Performance and Impact on Aircraft Noise
Aircraft design and operation are tightly coupled; near-term benefits in terms of both fuel consumption and aviation noise are possible by studying the interactions between aircraft configuration and operation simultaneously. Typical noise abatement procedures at airports, forcing aircraft to arrive and depart at airports following specific paths, likely decreases noise in certain communities, but increases fuel consumption and possibly noise levels in other surrounding communities. Similarly, when designing aircraft to meet stringent noise and emissions requirements, the operational costs and flight times likely increase. This research integrates aircraft design tools and flight simulations in order to find synergisms between the design of an aircraft and how it can be operated.