The D-optimality criterion states that the best set of points in the experiment
maximizes the determinant | X
T
X |. "D" stands for the determinant of the X
T X
matrix associated with the model. From a statistical point of view, a D-optimal
design leads to response surface models for which the maximum variance of the
predicted responses is minimized. This means that the points of the experiment will
minimize the error in the estimated coefficients of the response model.
The advantages of this method are the possibility to use irregular shapes and
the possibility to include extra design points. Generally, D-optimality is one of the
most used criteria in computer-generated design of experiments.
Several applications are described in Giunta et al. (1996) for the wing design of
a high-speed civil transport and Unal et. al. (1996) for a multidisciplinary design
optimization study of a launch vehicle. Haftka and Scott (1996) have reviewed the
use of D-optimality criteria for the optimization of experimental designs