Closed-form mechanical models to predict the behaviour of complex stru
ctural systems often are unavailable. Although reliability analysis of
such systems can be carried out by Monte Carlo simulations, the large
number of structural analyses required results in prohibitively high
computational costs. By using polynomial approximations of actual limi
t states in the reliability analysis, the number of analyses required
can be minimized. Such approximations are referred to as Response Surf
aces. This paper briefly describes the response surface methodology an
d critically evaluates existing approaches for choosing the experiment
al points at which the structural analyses must be performed. Methods
are investigated to incorporate information on probability distributio
ns of random variables in selecting the experimental points and to ens
ure that the response surface fits the actual limit state in the regio
n of maximum likelihood. A criterion for reduction in the number of ex
periments after the first iteration is suggested. Two numerical exampl
es show the application of the approach.