A stochastic analysis of structures usually requires multiple reanalysis of
the structures to compute the statistics of the structural response. A sim
ilar problem exists in the design of optimal structures where many analysis
are needed before reaching the extremal solution. In both fields the reana
lysis requirements are considered as an unacceptable numerical burden. To c
ircumvent the reanalysis obstacle investigators have been using approximate
analysis. Common methods are first and second-order series expansions of t
he nodal displacements, and related perturbation methods. Interestingly, a
popular approximation method in structural design, the reciprocal approxima
tion technique, has not been used in stochastic analysis. This paper shows
that this method can easily be used to compute the statistics of the struct
ural response. When expanding the displacements linearly in terms of the re
ciprocals of the element stiffnesses, one obtains, as a rule, better result
s than with a linear Taylor expansion. It is shown that for low values of t
he relative redundancy the method yields second order quality approximation
s. Unlike many other techniques, the reciprocal approximation also produces
the statistics of the internal forces. The theory is illustrated with typi
cal beam and arch trusses and is compared with existing stochastic methods.
(C) 2000 Elsevier Science Ltd. All rights reserved.