PROBABILISTIC PERFORMANCE-BASED DESIGN USING GA-BASED NONLINEAR RESPONSE OPTIMIZATION

摘要

Genetic and evolutionary algorithms (GA/EA) have been used extensively in recent years to solve complex optimization problems. Advancements in the field of structural engineering have resulted in the development of performance-based design (PBD) methodologies for buildings. Although at early stages in development, it is likely that PBD will see widespread practical implementation resulting from the common language used (e.g. performance expectation) by all stakeholders in the building process. The current research effort is geared towards development of an automated PBD environment for steel structural systems using GA's and nonlinear structural response analysis tools. The demand and resistance factor design (DCFD) methodology of FEMA-350 can be used to evaluate confidence levels associated with the probability of a structure exceeding a limit state (performance levels). The confidence levels obtained are then incorporated into a GA fitness function along with initial construction "cost" in a multi-objective design scenario. The GA, through the processes of simulated natural evolution, creates generations of solutions that eventually adapt to the design criteria (environment). In this study, these criteria are the minimum cost associated with maximum confidence in meeting structural seismic performance objectives of collapse prevention and immediate occupancy. Nonlinear response history analysis is used to evaluate seismic demand under different levels of hazard. The automated computational design platform was examined on several benchmark problems. The classical ten-bar truss was used to verify the compatibility of the DRAIN-2DX [1] structural analysis engine and the GA [2]. A portal frame with fully- and partially-restrained (FR, PR) connections is designed using two sets of seven earthquake ground motion records. Nearly equal weight designs for three frame and connection configurations are discussed. Conclusions regarding seismic performance, the optimality of designs, and nonlinear pushover response characteristics are made.