In real applications, it is extremely difficult (if not impossible) to define function(s) that measure the `true merit' of a design object. For example, even the most prominent aircraft designers would not dare to claim that a particular set of merit and constraint functions measures the `true merit' of a class of aircraft. The traditional approach to set up a `machine centric' optimization cannot effectively address this issue because there is no user interaction with the numerical optimization at design or run time. The semi-automatic concept, however, can help because the user is allowed to interact with the design problem and the design progress in-the-loop, such that the design criteria can be improved relatively easily at design time. This paper briefly describes the semi-automatic design optimization setup which was introduced in full detail in a previous paper by the author. A simple multidisciplinary aircraft conceptual design optimization problem is then specified based on Torenbeek (1992). Various modes of in-the-loop user control on the search progress and the search problem then illustrate the potential benefits of allowing the user to interact with a numerical agent at various levels of automation.

, , , ,
, , ,
CWI
Software Engineering [SEN]

Shahroudi, K.E. (1997). Aircraft conceptual design by collaborative manual and automatic agents. Software Engineering [SEN]. CWI.