Methods of evaluating structural reliability have advanced from member to system analysis. A nonlinear structural analysis program is utilized to derive ultimate and serviceability limit states in load space. System probability of failure is computed by integration of the joint density function of applied loads in the failure region in a three-dimensional load space. The reduction of load space dimensions to three is possible by assumption of positive perfect correlation amongst certain floor loads. Structural variability is modeled by randomization of the limit state function in load space without requiring any additional structural analysis. Sensitivity of failure probability to load and resistance variables are studied and factors with significant contributions are noted. A systematic load space approach to design, based on applying weights to modes of failure and minimization of the total expected cost using single and multiobjective linear programming, is developed. Results are trade-off curves between future (or failure) cost and initial cost for several realistic structures.
