Abstract We describe a framework that includes ASTROS for structural and loads analysis, object-oriented MATLABtools for reliability analysis, and DOT for optimization and most probable point estimation. These tools areused to perform minimum weight design of a ten-bar truss and a representative aircraft wing. Reliabilityconstraints include tip displacement, natural frequencies, and aileron effectiveness. The reliability analysisalgorithm uses adaptive nonlinear approximations to compensate for nonlinearity of the failure surfaces. Introduction Aerospace structural design traditionallyrelies upon safety factors to compensate for uncer-tainties in material properties, component dimensions, external loads, and imprecise mod-els. However, computational structural mechanics has evolved to a level of sophisticationthat merits consideration of newer methods for incorporating uncertainty in the design pro-cess. Multidisciplinary optimization (MDO) has developed in parallel with computationalanalysis; within this framework, several performance constraints must be satisfied simulta-neously, usually with the objective of minimizing structural weight. ASTROS (UAI, 1997)reflects the state of the art in automated optimization of aerospace structures; however,safety factors still are used to account deterministically for uncertainties.Future aircraft, to realize design and operational cost savings, will be designed with thegoal of certification by analysis. This objective will require more detailed considerationof uncertainty early in the structural design process. Probability theory forms the basis ofstructural reliability analysis, a reasonably mature approach to estimating the likelihoodthat a proposed design will satisfy the imposed performance constraints (Melchers, 1999).For complex structures that require finite element analysis, the constraints are implicit andtherefore computationally expensive to evaluate. Approximation techniques help to alle-viate this difficulty; in particular, the benefits realized through the use of adaptive approx-imations were demonstrated by Luo and Grandhi (1997). They extensively modified theASTROS executive control sequence to include reliability constraints in the optimizationprocess