Reliability-based design optimisation of geotechnical systems using a decoupled approach based on adaptive metamodels

Geotechnical engineering analysis and design involves considerable uncertainties. Reliability-based design optimisation (RBDO) intends to minimise an objective function such as material quantity or cost subject to design requirements specified by probabilistic constraints. In this work, the RBDO of geotechnical systems is studied using a decoupled approach based on adaptive metamodels. To improve the traditional double-loop structure of RBDO, the proposed approach decouples the numerical reliability analysis loop from the design optimisation loop. A metamodeling method based on augmented radial basis function (ARBF) is adopted to create approximate functions of the reliability indices, so that the design optimisation is performed based on the metamodels of the probabilistic constraints. To improve the accuracy of predicting the reliability indices, an adaptive technique of the ARBF metamodels is applied. The optimal point in one design iteration is treated as an additional sample point for updating the metamodels of the reliability indices. The accuracy of the metamodels is progressively enhanced through this dynamic process, especially in the neighbourhood of the optimal point. Mathematical and geotechnical engineering examples are solved and numerical results are presented. The proposed design optimisation framework works well and is a useful alternative for solving RBDO of geotechnical engineering problems.

Automated summary

This research introduces a new RBDO approach for geotechnical systems, which separates numerical reliability analysis and design optimization using adaptive metamodels, achieving promising results.