A metamodeling-based robust optimisation approach for structures subjected to random underground blast excitation

Abstract

The blast hazards are associated with extreme randomness. Thus, structures under blast should be properly designed for such highly uncertain phenomena. However, codal stipulations and existing literature do not consider important uncertainty-related design issues, such as record-to-record variation of blast load, high uncertainty levels associated with blast parameters, and robustness of structural behaviour under extreme blast events. In the present study, a computationally efficient robust optimisation (RO) approach is proposed for blast-excited structures by addressing these issues. In doing so, a new CDF-based constraint formulation is developed in the dual response surface-based metamodeling framework to avoid the use of direct Monte Carlo simulation (MCS) during RO iterations. Thereby, the proposed approach becomes computationally much efficient than the conventional direct MCS-based RO. Since the proposed approach resorts to matching the CDF of the actual response, instead of the conventional RO formulations based on the first two statistical moments only, the approach is more realistic. The present approach adopts the moving least-squares method instead of the usual least-squares method to retain accuracy in response approximation. The blast load time-history is generated by a stochastic field model with proper validation through real blast data. Three example problems have been investigated to establish the effectiveness of the proposed RO procedure. The RO results depict that the present method yields economic design solutions, which are also insensitive to high uncertainty levels associated with the extreme blast load. The COV of performance function is substantially low even when input parameter uncertainty is high.