A sequential reliability assessment and optimization strategy for multidisciplinary problems with active learning kriging model

Abstract

The reliability-based multidisciplinary design and optimization is of significance for increasing the quality and economic efficiency in many industrial designs. However, the intensive coupled multidisciplinary analysis and reliability assessment make it impractical for real engineering problems due to the unacceptable computational cost. In this paper, we studied different active learning kriging (ALK) models to approximate the objective function for optimization purpose and the limit state function for reliability assessment purpose. Then, we proposed a sequential reliability assessment and optimization strategy, which satisfies the reliability constraints and then moves to the current optimal point to save the computational resource for optimization iteration, to conduct efficiently the reliability-based multidisciplinary design and optimization. Combined with modified ALK models, the computational cost could be saved by this strategy and it could be demonstrated with a practical case of reliability-based multidisciplinary design and optimization for twin-web turbine disk involving fluid, thermal and structural disciplines.