| description abstract | Abstract. Traditional reliability-based design optimization (RBDO) methods are typically developed to address problems with a single-target reliability. However, as the complexity of engineering structures increases, there are cases where the reliability requirements cannot be predefined, or where multiple reliability levels are required. To find the optimal RBDO solutions for different target reliabilities, multiple surrogate models need to be constructed, resulting in extremely high computational burden. To address this issue, a new multitarget reliability-based design optimization (MTRBDO) method is proposed to efficiently and accurately identify different RBDO optima considering multiple target reliabilities simultaneously. Treating the probability of failure as an additional objective function alongside the original objective function, the RBDO problem with multiple target reliabilities is converted into a multi-objective design optimization (MODO) process. A two-stage surrogate modeling strategy is developed to efficiently create surrogate models of failure probability by adaptively sampling in the insufficiently fitted vicinity of the limit state functions and calibrating the model within target failure probability interval. The constructed surrogate model is used in the MODO procedure, eliminating the need to repetitively create multiple surrogate models for different RBDO procedures corresponding to different target reliabilities. Moreover, an improved nondominated sorting genetic algorithm II (NSGA-II) is further developed to obtain the Pareto front of the created MODO process. The MTRBDO optima corresponding to a series of target reliabilities are directly obtained from the Pareto front. The proposed MTRBDO method is tested on a numerical example, a wind turbine tower design case, and an engineering application of tunnel boring machine cutterhead. Results demonstrate that the proposed MTRBDO method can more efficiently find multiple RBDO optima compared with existing RBDO methods and reveal the relationship between optimal design and reliability. | |
