ABSTRACT
Uncertainties inadvertently present in material properties, environmental, loading, and geometry conditions of a structure. Due to the presence of dynamic uncertainties, ensuring the time-dependent reliability (TDR) of the structural design during reliability-based design optimization (RBDO) is of critical importance. Continuous material degradation and the aging process of a structure introduce a new challenge to the RBDO problems. An integrated framework is proposed to optimize a structural design while predicting the TDR of a structure in this work. The dynamic limit state functions (LSFs) are considered as constraints to the RBDO problem. A framework integrating transfer learning and Kriging surrogate model is proposed to predict the TDR. This research works attempts to predict the TDR in the future time by utilizing only the LSF information available for a partial interval. Stochastic process samples (SPS) at different times as represented as similar using a transformation matrix (TM), which is calculated using transfer learning. The reliability is then predicted for the complete time interval using the same surrogate model. Once the reliability is predicted, the method then utilizes a gradient-based algorithm to find an optimal design of the structure. The proposed method's effectiveness is validated using various examples.
