Certifiable Risk-Based Engineering Design Optimization

Reliable, risk-averse design of complex engineering systems with optimized performance requires dealing uncertainties. A conventional approach is to add safety margins a that was obtained from deterministic optimization. Safer designs require appropriate cost and constraint function definitions capture the risk associated unwanted system behavior in presence The paper proposes two notions certifiability. first based on accounting for magnitude failure ensure data-informed conservativeness. second ability provide optimization convergence guarantees by preserving convexity. Satisfying these leads certifiable risk-based (CRIBDO). In context CRIBDO, measures superquantile (aka conditional value-at-risk) buffered probability are analyzed. CRIBDO contrasted reliability-based (RBDO), which uncertainties accounted via through structural thermal problem. reformulation short column problem leading convex presented. formulations more information about assign conservativeness, exhibit superior properties underlying functions, alleviate adverse effects choosing hard thresholds required RBDO.