Reliability-based design optimisation of structural systems using high-order analytical moments

•A novel moment-based reliability-based design optimisation framework is proposed.•An analytical high-order moment-based uncertainty evaluation technique is employed.•Genetic programming and selective sampling technique is used for metamodeling.•Fast and accurate reliability estimation achieved through high-order moments.•The proposed moment-based optimisation is more robust than the mainstream methods. Reliability-based design optimisation paradigm has become increasingly popular to achieve economical yet safer structural designs. However, within the iterative optimisation procedure, it is a challenging problem to simultaneously satisfy both, accuracy and computational efficiency of reliability analysis, particularly for nonlinear or large design problems. Addressing the shortcomings of the mainstream reliability estimation methods, this paper presents a new reliability-based design optimisation method that combines an analytical high-order moment-based uncertainty evaluation with an efficient response surface modelling. The proposed framework allows for the precise calculation of reliability through accurate high-order moments. The fast and accurate reliability estimation in combination with efficient sampling, in turn, reduces the total number of finite element analysis in achieving the final design. The proposed moment-based design optimisation methodology was tested on problems ranging from the design of a simple cantilever to a three-dimensional multistorey steel structure. It outperforms the mainstream methods with higher accuracy and lower computational burden especially when applied to a highly nonlinear numerical problem.