An efficient nonlinear stochastic finite-element method is proposed fo
r the design of frame structures, specifically applicable to steel str
uctures. The assumed stress-based finite-element method is used to com
pute nonlinear structural responses;and the corresponding response gra
dients. Nonlinearities due to geometry, material, and partially restra
ined (PR) connections are considered in the algorithm. Elastoplastic m
aterial behavior is considered and the PR connections are modeled by t
he four-parameter Richard model. The material properties, geometry, co
nnections parameters, and external loads are considered to be random v
ariables. Both the strength and serviceability limit states are consid
ered. The efficiency, robustness, desirability, and application potent
ial of the method are demonstrated by two examples. Several observatio
ns with design implications can be made from these examples. Frames de
signed for strength may not be safe when serviceability is considered.
The presence of PR connections makes the frame more flexible, and ser
viceability could become the governing limit state. It is essential to
properly consider the PR connections in the analysis and design of fr
ames, since design forces change significantly. In many cases, the mem
ber sizes may change considerably. The assumption that the connections
are rigid, which is routinely used in the profession, is not appropri
ate. The proposed algorithm can be used as an alternative to the curre
ntly available methods to analyze and design frames, and to evaluate t
he corresponding reliabilities. The design is expected to be more effi
cient using this approach.