A Novel Approach to Efficient Risk-Based Optimization

Optimization under uncertainties is a computationally demanding problem, due to the nested structural analysis, reliability analysis and optimization loops. While many shortcuts have been proposed and developed for reliability-based design optimization, no such shortcuts where available until today for risk-based optimization. In reliability-based design optimization, failure probabilities are constraints, which sometimes can be replaced by deterministic constraints. In riskbased optimization, failure probabilities are part of the objective function. Due to this fundamental difference, shortcuts for solving reliability-based design optimization problems do not apply to risk optimization. In this paper, a novel approach to solving structural risk optimization is presented. This approach leads to several-fold reduction in the computational effort for solving risk optimization problems. For each directional (line) search, a couple of Monte Carlo samples is used to obtain critical values of the design variables, leading to a complete description of the expected cost of failure function. Optimization algorithms can then be used for solving for the global optimum. The total cost of the optimization solution is equivalent to a single solution of the reliability problem by means of Monte Carlo simulation. The technique presented in this paper opens new possibilities for the solution of realworld large numerical structural optimization problems.