MDAO


What is MDAO?

MD(A)O stands for multidisciplinary design (analysis and) optimization. Basically, MDAO involves using computer simulations and mathematics to model, analyze, and semi-automatically design, the best possible systems. “Systems” might include aircraft, spacecraft, buildings, or really any entity that can be modeled and where we care about performance.

M (Multidisciplinary) - This means that multiple analysis disciplines are involved in the simulation or design process. Examples of disciplines might include aerodynamics, structures, weights, stability & control, and cost / finance.

D (Design) - The ultimate goal of MDAO is to produce good designs. In a broad sense, the design of a product is the input to an analysis code, and the outputs are metrics which tell us how well the design performs and whether it is feasible.

A (Analysis) - Many people in the field will use “A” in MDAO to explicitly state that all MDAO codes perform some kind of analysis. When a simulation is used to analyze a design rather than perform numeric optimization, the term MDA can be used. My lab at Michigan prefers to simply use MDO since optimization implies that analysis is being performed.

O (Optimization) - Broadly used, the word optimization simply means the process of making something better. In the context of MDAO, optimization refers to a set of specialized mathematical techniques that are used to find the “best” (optimal) design possible that meets a set of requirements. It is typical in MDAO (especially where geometry is a design variable) to use the techniques of constrained nonlinear optimization (or nonlinear programming).

The header image of my site is a 2D rendering of the 3D Rosenbrock Function (also known as the “banana function”) which is a challenging test case for an optimizer because it has subtle valleys to find.

Some challenges in MDAO include:

  • Computational cost - Aerodynamics and structural codes take a long time to run, even on supercomputers.
  • Fidelity - How precise and accurate are the computer models? How well does the simulation capture all relevant effects?
  • Flexibility - Some MDAO codes are only applicable to a narrow range of designs or conditions.
  • Globality - Some optimization techniques only incrementally improve a starting design and miss the very best point somewhere else in the design space; no generally useful technique guarantees that the best global solution is found.
  • Ease of use - Many MDAO codes have a very steep learning curve and are only useful to specialists. Acceptance - The interdisciplinary nature of MDAO means that not all the relevant experts may agree on the results of an MDAO study.
  • Visualization - MDAO design spaces are typically in many dimensions - more than 2 or 3. This makes conveying results difficult.